Paper Radar · today in AI

★ {\tau}-bench-LILT: Evaluating Conversational Agents on Realistic Multilingual Tool-Use Tasks

Kaden Uhlig, Kiarash Mokhtari, An{\i}l Arslan, Yunsu Kim, Joern Wuebker — Fri, 06 Nov 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 96%

Frontier evaluations of tool-using conversational agents are conducted almost exclusively in English, leaving a critical gap for organizations deploying agents in other languages and regions. We present ${\tau}$-bench-LILT, a multilingual extension of ${\tau^3}$-bench covering German, Korean, Turkish, and Farsi in the airline, retail, telecom, and banking domains. We natively localize every layer of the benchmark: customer databases are populated with culturally appropriate records, user-simulator prompts are refined iteratively through contrastive annotation by native-language experts, and tasks are updated according to the localized database and to reflect realistic tasks in the target region. Evaluation across eight agent models shows the largest drops compared to English in the Korean and Farsi retail domains, by up to 61 points in pass^4. Code and data released at https://github.com/lilt/tau-bench-lilt.

★ MCPHallu: Benchmarking Reasoning, Execution, and Memory Hallucinations in MCP Agents

Yizhen Jiang, Haowei Guo, Tianyi Bai, Zengjie Hu, Lu Ma, Zhengyang Zhao — Fri, 06 Nov 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 97%, Hallucination 94%

Large language model agents are increasingly deployed through the Model Context Protocol (MCP), which allows them to discover and invoke tools across multiple servers at runtime. Existing MCP benchmarks primarily measure task completion; where they diagnose failures, they do so after execution rather than building specific reasoning, execution, and memory failure risks into task construction. We introduce MCPHallu, a benchmark for diagnosing hallucination-prone agent failures in MCP environments along three axes: reasoning, execution, and memory. MCPHallu contains 358 tasks across five capability domains and runs on 36 production MCP servers in isolated containers. Each task is designed with a structural trigger that makes one of four subtypes the dominant failure risk: Branch Collapse, Unreachable Goal, Tool Misuse, or Context Forgetting. This design enables subtype-level analysis rather than aggregate task success alone. We evaluate fourteen frontier LLM agents from eight vendors on 5{,}012 trajectories. Current MCP agents remain vulnerable to hallucination-inducing tool environments, achieving an average score of 0.642 on a $[0,1]$ scale (higher is better). Failures concentrate on Unreachable Goal tasks, where the average score drops to 0.458 and 23.2\% of runs pass, indicating that agents often continue issuing tool calls or claim successful completion rather than recognizing infeasible goals. Subtype rankings diverge sharply: a model's rank shifts by 6.9 of 13 positions on average across subtypes, and Branch Collapse is uncorrelated with Unreachable Goal across models ($r = -0.12$). MCPHallu is released at https://github.com/J1zz/mcp_hallu.

★ TCSExam: Evaluating Frontier LLMs on the Full Range of TCS Research

Yifan Luo, Zhengying Liu, Yang Yuan, Zhilin Yang, Andrew Chi-Chih Yao — Fri, 06 Nov 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 31%

Can large language models (LLMs) advance theoretical computer science and resolve its long-standing open conjectures? To study progress toward this goal, we introduce TCSExam, a unified framework and benchmark covering the full range of TCS research. It evaluates theorem proving, solver engineering, and constructions alongside direct attempts to resolve open conjectures, capturing both concrete research accomplishments and progress toward long-standing research goals. TCSExam-Full contains 100 expert-reviewed tasks, with submissions checked by the Lean kernel or an executable task checker. TCSExam-Lite selects 20 tasks, five per track, reducing the number of task attempts by 80% at the same per-attempt budget. We evaluate eight agents on Full and nine on Lite. The TCSExam Index combines all four tracks: GPT-6 Astra leads Full (71.9), and Opus 5.5 leads Lite (72.3). Full and Lite rankings are strongly aligned across eight shared agents (Spearman's $\rho=0.93$), supporting Lite as a compact, discriminative benchmark. The evaluations yield improved solver outputs and mathematical constructions, together with independently verified special cases and bounds from open-conjecture attempts. TCSExam provides an extensible framework with task sources, Docker build recipes, and evaluation tools for reproducible use.

★ From Twisted Facts to Rogue Agents: Can LLMs Reliably Detect Multi-Dimensional Hallucinations?

Hanyu Qiu, Ze Liu, Zhe Zhao, Kuang Xiong, Blaze Chen, Zhenxiang Yan — Fri, 06 Nov 2026 00:00:00 +0000

99% · Evaluation 99%, Hallucination 97%, Agents 51%

Large language models (LLMs) are widely used for information access, content production, and agent-based task execution, but their outputs remain vulnerable to subtle, high-risk hallucinations. To systematically evaluate and advance post-generation hallucination detection, we introduce HaluCheck, a large-scale, bilingual benchmark comprising over 42k instances spanning both factual hallucinations in text and faithfulness hallucinations in agent trajectories. HaluCheck features fine-grained, multi-level annotations for detection, localization, and correction, covering different tasks and perturbation degrees. Experiments on fourteen LLMs reveal that, while models show basic sample-level detection ability, they struggle with fine-grained localization, correction, and agent hallucination detection. Further analysis demonstrates that thinking, web search, and supervised fine-tuning each offer partial improvements, but none fully solve the problem. HaluCheck thus provides a solid basis for benchmarking and diagnosing LLM reliability across text and agent scenarios.

★ TriAttention: Unified Sparse Mixed-Precision Attention for Long-Context LLM Inference

Hanzhi Zhang, Qiao Zhang, Qinglei Cao, Heng Fan, Yan Huang, Kewei Sha — Fri, 06 Nov 2026 00:00:00 +0000

98% · Efficiency 98%, Evaluation 46%

Sparsity and quantization reduce the cost of attention by skipping interactions and lowering the precision of those retained. However, useful interactions differ in their precision requirements: some tolerate low-precision scores while remaining sensitive to removal. We therefore introduce TriAttention, which assigns query-key (QK) tiles to skipping, 4-bit, or 8-bit computation, using 4-bit to retain quantization-tolerant interactions at lower cost. To represent compactly, we encode each head's computation regions with two diagonal-distance boundaries, keeping policy storage independent of context length. Offline calibration then evaluates candidate boundary pairs using normalized projected-output errors and complete-tile execution costs. Using these measurements, a model-wide allocator selects each head's boundaries to minimize calibration error under a shared computation budget. During prefill, the selected boundaries generate precision-grouped traversals for a fused kernel that skips omitted tiles and combines contributions from both QK precisions into one normalized attention output. The quantized keys and values prepared during prefill are then reused in a compressed key-value (KV) cache for decoding, reducing repeated reads of prompt data. Experiments show that TriAttention achieves end-to-end generation speedups of up to 7.88x over the Torch attention and 2.65x over the BF16 baseline. Meanwhile, TriAttention maintains quality comparable to the original models on general-purpose and long-context tasks. These results show how a compact retention-and-precision policy translates local sensitivities into attention computation and repeated KV-cache access. The implementation is https://github.com/HanzhiZhang-Ulrica/TriAttention.

★ FINCOSMOS: An Open-World Evaluation Paradigm for Agents in Finance

Yupeng Cao, Xueqing Peng, Lingfei Qian, Haohang Li, Yangyang Yu, Anke Xu — Fri, 06 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 96%, Hallucination 78%

Language agents increasingly work for users on the open Web, and finance is where trusting them is hardest. The data must come from the Web, the strategy lives in the user's head, and correctness depends on time, so an agent can fail silently: a strategy the user never asked for, unverified data, or a backtest that uses the future yields a report as convincing as a correct one. Existing benchmarks close the world somewhere, fixing the task, the data, or the timeline, and thereby remove these failures instead of measuring them. We introduce FINCOSMOS, an open-world evaluation paradigm for agents doing financial work. A simulated user holds a hidden strategy sampled from a compositional grammar and reveals it only when asked. The agent receives no data, so it must acquire data from the Web, build its own backtest under a data cutoff, and deliver verifiable results. Grading needs no reference answer. Programs verify point-in-time correctness and recompute every reported metric, the delivered strategy is checked against the hidden specification, and a run is accepted only when the user is satisfied and none of 15 veto checks fails. Across four coding agents and a model without tools in 150 sessions, no session is accepted: every session that the simulated user signed off fails a program check, and agents raise only 54--64\% of the details they were not told. Human experts attribute 88.33\% of failures to the agents. With FINCOSMOS feedback, an agent with persistent memory learned all ten tasks.

★ CST-Bench: Benchmarking Counterfactual Spatial Transition Reasoning with Causal Transition On-Policy Self-Distillation

Ying Li, Yongle Cheng, Jing Li, Xingquan Cai, Yawei Luo — Fri, 06 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, New model 88%

Spatial reasoning is a fundamental capability of multimodal large language models (MLLMs), yet existing benchmarks mainly focus on static scene understanding. When objects are moved or swapped, models must infer not only spatial relations but also how interventions change the scene state. To address this problem, we introduce CST-Bench, a benchmark for counterfactual spatial reasoning. CST-Bench contains 7,082 examples covering 6 intervention types and 12 transformation operators, with dense causal annotations including intervention sources, affected relations, expected effects, invariant relations, and metric changes. We benchmark general-purpose MLLMs on CST-Bench and further compare post-training strategies on CST-Bench and standard spatial benchmarks. Results show that methods effective for ordinary spatial reasoning can still fail under counterfactual interventions. We further formulate counterfactual spatial reasoning as causal state-transition modeling and propose CT-OPSD, which uses CST-Bench annotations to guide a teacher model in learning counterfactual state transitions and causal reasoning chains. These chains are then distilled into a student model that reasons only from real observations. Overall, experiment results establish CST-Bench as a challenging benchmark and reusable supervision resource for counterfactual spatial reasoning. CT-OPSD further provides an initial demonstration of how its annotations can support model improvement.

★ Binding Paraphrased Instructions to Encountered Objects in sLLM Agents under Partial Observability

Fri, 06 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Efficiency 51%

Reliable deployment of LLM agents requires grounding diverse user expressions in their environments. Although RL-finetuned smaller large language model (sLLM) agents rival larger prompted LLMs on embodied tasks, their robustness to instruction paraphrases remains unclear. We introduce ALFWorld-Para to evaluate this robustness under controlled paraphrases, revealing 21-52% declines in success rates (SRs), predominantly driven by object synonyms. Despite retaining knowledge of synonym referents, agents often fail to take the corresponding objects even after observing them, exposing a gap between referential knowledge and object binding. Partial observability further complicates binding, as referents may remain unseen or have already been encountered under different labels. To address this challenge, we propose BINDIntent, a training-free framework that probabilistically binds referring expressions to observed labels using semantic proximity and estimated persistence against unseen candidates. Without additional training or LLM calls, BINDIntent improves synonym-paraphrase success of RL-finetuned sLLMs by 17.5-49.1 percentage points (pp) with only 0.4 ms of CPU overhead per step. It also raises GiGPO-7B's overall SR from 61.8% to 77.4%, surpassing a prompted 24B LLM. Code and benchmark are available at https://github.com/cau-hai-lab/ALFWorld-Para.

★ Measuring the Marginal Value of Verifier-Based RL Beyond Supervised Fine-Tuning

Amit Sagtani, Saloni Garg — Fri, 06 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

Reinforcement learning from verifiable rewards (RLVR) is usually evaluated as an alternative to supervised fine-tuning (SFT), or as a stage that follows it, with the implicit assumption that the RL stage contributes gains of its own. We measure that contribution directly, in a controlled setting where both stages can be run to completion and compared exactly. We built gpu-train-debug, a deterministic benchmark of 1,877 synthetic GPU and LLM training failure diagnosis tasks across 30 archetypes, graded by a programmatic verifier with partial credit, and calibrated so that a mid-tier model scores in a 0.60 to 0.75 band. A pilot produced a surprise: 1,200 gold demonstrations nearly solved the benchmark (SFT 0.785 on the hard split), while 15 steps of GRPO from the base model did not move it (0.314 vs 0.311 base). We therefore amended the study, in writing and before the full run, to ask the sharper question: what does verifier-based RL add beyond demonstrations? At the pre-registered scale (60 GRPO steps, learning rate 1e-6), the answer is: no detectable improvement in-distribution (+0.22 points, 95% CI [-0.06, +0.51]), and a modest but measurable improvement under distribution shift (+0.33 points on a held-out OOD split, CI [+0.02, +0.68]). SFT accounted for most of the available improvement, while verifier-based RL produced only modest additional gains under distribution shift. The residual errors SFT leaves are mostly systematic, identical across samples, which likely reduces the within-group learning signal available to GRPO. A full 60-step cold-start GRPO arm confirms the pilot at the full horizon. The whole study cost $8.69 of compute and is fully reproducible from archived generations, frozen configs, and tagged code.

★ Circe: Training-Free LLM Self-Evaluation via Component Interaction Consistency

Shuyi Miao, Zheng Li, Xu Xie, Linrui Xu, Fei Shen, Xiaoyu Du — Fri, 06 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 72%

As large language models (LLMs) are increasingly deployed in real-world applications, evaluating the correctness of their responses is crucial for reliable use. However, such evaluation is challenging because reference answers are rarely available in practice. Existing reference-free methods mainly rely on output signals, which may assign high scores to incorrect responses and thus do not necessarily reflect correctness. This motivates exploring the model's internal computation for complementary correctness signals. Recent studies exploit internal states, but they mainly analyze hidden-state trajectories across layers, which record only the combined updates of attention and feed-forward networks (FFNs) and overlook how these components interact. We therefore propose Circe, a training-free self-evaluation method that assesses a response using both internal and external signals, without reference answers or repeated sampling. For internal signals, we first introduce component interaction consistency (CIC), which measures the directional alignment between attention updates in earlier layers and attention-induced FFN output changes in later layers. We further incorporate representation geometry (RG) to characterize normalized displacement and angular changes along layerwise hidden-state trajectories. For the external signal, we convert the mean negative log-likelihood of generated tokens into a bounded likelihood confidence (LC) score. Finally, we fuse CIC, RG, and LC into a unified correctness score. Experiments on five LLMs and six benchmarks across four domains show that our method distinguishes correct from incorrect responses more effectively than existing baselines, and generalizes to multilingual, multimodal, and open-world settings.

★ Beyond the Keyword: A Human-Annotated Benchmark for Arabic Anxiety-Related Discourse

Fri, 06 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

Keyword retrieval returns expressions of personal anxiety together with many other uses of anxiety-related language. We present a benchmark of 1,000 Arabic Twitter/X posts retrieved with the keyword qalaq ("anxiety/worry") and annotated for six discourse functions. Two native Arabic-speaking annotators labeled every post independently, reaching 51.1% raw agreement (Cohen's $\kappa = 0.378$). We then reviewed all 489 disagreements and all 511 agreements under the same guidelines. The review changed 120 of the agreed labels (23.5%), which shows that agreements need auditing as well as disagreements. The final labels form a consensus reference; on a random sample of 200 posts, an additional annotator agreed with it on 60.5% of posts ($\kappa = 0.517$). Against this reference, Fanar-C-2-27B reaches 79.8% accuracy and macro-F1 0.784 ($\kappa = 0.749$). Its errors concentrate in the residual category (41.1% of errors; precision 0.905, recall 0.447), whose definition is internally inconsistent in the executed prompt, so the low recall cannot be attributed to the model alone. Sociopolitical Anxiety (27.1%) outnumbers Personal Emotional Expression (22.0%); these are assignment proportions in the sample, not estimates of prevalence or retrieval precision. We release the guidelines, a residual-subtype scheme, a fixed data split, label provenance, and evaluation code.

★ Terminal-style Mathematical Modeling Bench: The First Agentic and Verifiable Benchmark for Mathematical Modeling

Xiaojun Wu, Cehao Yang, Xuhui Jiang, Chengjin Xu, Jia Li, Jian Guo — Fri, 06 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 46%

Mathematical modeling turns an open-ended real-world problem into assumptions, a model, a solution, and a defended recommendation, and its answers matter only if they are both close to a credible reference and physically feasible; existing benchmarks either score solver fragments or grade reports with an LLM judge, leaving those numbers unchecked. We introduce TB-MathModel (the Terminal-style Mathematical Modeling Bench), the first benchmark to make complete contest-style modeling problems agentic, verifiable, and terminal at once, over 18 real CUMCM and MCM problems (2023--2025). Each task ships the verbatim statement, official attachments, and figures in a containerized terminal; the agent returns one JSON artifact, scored on the numeric core of the final question against a reproduction of the Outstanding-award paper. Two design choices distinguish it: direction-aware, one-sided scoring, which never penalises beating the anchor, and two-layer feasibility verification---a programmatic gate over six decidable invariant forms plus 624 Lean~4 obligations---with a hard gate that zeroes infeasible answers. Each task is scored against a panel of human outstanding papers rather than one, so a model must be close to many accepted answers; across seven multimodal entries (three seeds each) the strongest reach a multi-reference O-Eval of 64.58\%, while the \$2.54 open model holds 59.81\% at $1/100$ the cost. TB-MathModel turns the numeric core of modeling into a reproducible, machine-checkable signal for measuring long-horizon modeling competence and for collecting agent trajectories. The source code is available at https://github.com/wxj630/terminal-bench-math-modeling.

★ LeanContext: Disentangling Mathematical Resources from Local Formal Context in Lean Whole-Proof Generation

Hang Zhang, Qihao Ye, Litong Liu, Yunbei Zhang, Yugyung Lee, Yue Zhou — Fri, 06 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 30%

Neural theorem proving increasingly uses retrieved premises and project context to support proof generation. Evaluating these resources requires measuring the advantage of target-selected declarations over comparable alternatives and how this advantage depends on training. We introduce LeanContext, a benchmark of 129,826 Lean whole-proof generation problems, to investigate these questions. We group selected complete preceding mathematical declarations as Mathematical Context (MC) and the remaining selected local source commands as Environment Context (EC). Four aligned input views support crossed training and inference evaluations with shared targets and original Lean verification environments. For both diffusion- and autoregressive-based models, selected declarations outperform comparable alternatives by a larger margin when training includes MC. Final-input audits identify omitted declarations referenced by the target. Prioritizing these dependencies improves Pass@32 by 1.40% and 1.70% for LLaDA and Llama, respectively, on 2,000 targets. The selection experiment operates without access to target reference proofs and holds checkpoints, EC, and inference budgets constant. The code and dataset are available at https://github.com/haaaardcvZ/LeanContext.

★ Benchmarking Large Language Models on Heterogeneous Cognitive and Biomedical Knowledge Graphs: The Role of Graph Grounding in Multi-Hop Reasoning and Hallucination Mitigation

Ali Sarabadani, Kheirolah Rahsepar Fard, Tahereh Jeirani — Fri, 06 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 98%

Large language models (LLMs) are increasingly used in biomedical and cognitive neuroscience, but their ability to reason reliably over heterogeneous knowledge graphs (KGs) remains unclear. We benchmark five LLMs (GPT-5.1, Gemini 2.0 Pro, Llama 3, Qwen, and BioGPT) across five KGs: DrugBank, NeuroSynth, Cognitive Atlas, Brain Map, and GDPKG. We introduce a multitier query dataset to evaluate reasoning depth and factual accuracy under KG-aware (graph-grounded) and KG-agnostic conditions. Inter-annotator agreement was substantial, with Cohen's kappa of 0.781 (range: 0.753-0.807) and Fleiss' kappa of 0.77. Results show that general-purpose models, particularly GPT-5.1 and Gemini 2.0 Pro, generally outperform domain-specific models on multi-hop and high-level reasoning tasks, although performance declines as query complexity increases. KG grounding reduces hallucination rates across all evaluated models. Contrary to our initial hypothesis, domain-specific models do not consistently outperform general-purpose models on specialized tasks without graph grounding. These findings highlight the value of combining broad language-model capabilities with structured domain knowledge. We provide practical guidance for selecting LLM-KG integration strategies based on task complexity and computational constraints, and discuss hybrid neurosymbolic architectures for personalized medicine, cognitive disorder diagnosis, and explainable AI.

★ NPCBench: A Clinical Apprenticeship Benchmark for Multimodal Care-Pathway Reasoning in Nasopharyngeal Carcinoma

Pengkai Wang, Wei-Wei Zhang, Yan Li, Min Tang, Zhitian Hou, Zeyu Liu — Fri, 06 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

Current medical benchmarks typically break down clinical competence into isolated factual questions, single-image tasks, or static case vignettes. These tests score individual answers, but reveal less about whether a model can integrate subspecialty knowledge with multimodal evidence and track a patient's condition throughout a course of care. To close this gap, we introduce NPCBench, a clinical apprenticeship benchmark for nasopharyngeal carcinoma (NPC). NPCBench mirrors specialist training, from guideline and foundational knowledge to case planning, longitudinal review, CarePath integration, and open-ended consultation. It contains 4,777 component and consultation items and a 100-patient CarePath cohort with 417 scored diagnostic and care tasks. The episode-level evaluation tests whether diagnostic and care answers are jointly correct within a care pathway. By linking stepwise tasks to patient-level care trajectories, NPCBench evaluates whether models ground decisions in visual evidence, interpret changes in patient state, and meet clinical requirements in consultation. Across 30 evaluated models, we observe a composition gap: strong performance on individual tasks does not ensure jointly correct diagnostic and care answers for the same patient. Among 21 multimodal models, strict CarePath success reaches at most 23%. Paired controls and physician calibration further examine evidence grounding, longitudinal revision, and consultation judging. Overall, NPCBench broadens the evaluation of medical artificial intelligence beyond static question-answering paradigms to an apprenticeship-oriented assessment of clinical workflows, thereby offering a testbed for evidence-based and temporally coherent reasoning in oncology care.

★ XStack: A Benchmark for ML Repo Migration

Linrong Cai, Taiming Lu, Haozhe Chen, Zhuang Liu — Fri, 06 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 58%

ML code should run where researchers choose, not where it was first written. Today, machine learning code mainly lives in two ecosystems, PyTorch on NVIDIA GPUs and JAX on Google TPUs. Migrating repositories between them means building and testing across frameworks and hardware. This migration is (1) a practical need, since it unlocks different hardware for researchers, and (2) a natural benchmark: it involves different frameworks and hardware, covers whole repositories, and is long-horizon and challenging. We introduce XStack, a benchmark that evaluates agents on migrating ML repositories between these two ecosystems. Each agent receives a PyTorch repository and a clean TPU machine and must deliver a JAX repository. Its output passes if it meets our grading criteria, based on the source repository's published result, authenticated and calibrated by us. Migration by agents remains unreliable: Claude Opus 5 passes 75%, GPT-5.6-sol 45%, and every other agent at most 40%. XStack is a step toward freeing ML code from its ecosystem. We release XStack to make progress toward reliable migration across frameworks and hardware measurable.

★ LowRAM: Low-Rank Keys and Amortised Values for Composable KV Cache Compression at Test-Time

Teresa Delgado, Hugo Hazard, Martin Benfeghoul, Jun Wang, Zafeirios Fountas — Fri, 06 Nov 2026 00:00:00 +0000

97% · Efficiency 97%

Autoregressive decoding relies on the KV cache, whose memory footprint grows linearly with context length and caps batch size and throughput. Existing post-training methods exploit redundancy across tokens, layers, and feature dimensions, or lower numerical precision. We identify a complementary axis: within each attention head, values are largely predictable from their corresponding keys. We exploit this with value cache amortisation, which fits a lightweight per-head predictor at test time and stores residuals only for the tokens it reconstructs worst, giving $4\times$ near-lossless value compression. The predictor can be initialised by meta-learning for substantially reduced prefill overhead, or fitted from scratch in settings that allow a one-off cost, such as multi-query inference. We pair value amortisation with cross-layer low-rank key compression to produce LowRAM, an asymmetric compression pipeline that stays close to lossless across three model families. By reconstructing only a selected subset of KV pairs at each decoding step, LowRAM can keep decoding overhead minimal and achieve up to $5\times$ higher peak decode throughput at $4\times$ compression. LowRAM further composes with quantisation and eviction, reaching $26.9\times$ compression at $94.5\%$ of baseline performance on long contexts. These results show that treating keys and values as structurally distinct enables efficient, aggressive, and composable KV cache compression.

★ Closed-Loop Agent Evaluation: Layered Measurement and Self-Calibration

Ruifeng Ren, Gun Ahn — Fri, 06 Nov 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 42%

Once an agent is live, teams want to know whether it is doing its job, not just whether it throws errors. Yet most evaluation is one-off: a benchmark run, a look at traces, a single score. Its conclusions neither check themselves nor reach the agent. We propose closed-loop evaluation, an evaluation method for deployed agents. It models evaluation as performance estimation under biased raters and censored outcomes, and it requires every level's conclusion to trace back to atomic data and to be computed by versioned rules. The method covers four levels of analysis (span, trace, conversation and engagement), measures along a spine of Execution, Substance, Trajectory and Results, treats Conduct as a constraint throughout, and uses Provenance and Adjudication components to certify the evidence and the evaluators. Evaluator agents generated by an Agent Factory carry these components, each with a domain knowledge pack and an adapter harness. Two loops close the evaluation: an inner loop audits the evidence, rules and metrics themselves, and an outer loop uses per-component regression checks to drive agent improvement. We also give a pipeline that runs this process automatically. Using the business-analysis agent SharpInsight as a case and porting the method to public conversation data, we show how problems differ across levels and how evidence-and-rule judgment is more reliable than holistic scoring (ICC 0.86 vs. 0.81; kappa 0.81 vs. 0.63).

★ Recursive Sparse Attention: Million-Token Length Extrapolation at Logarithmic Cost

Ruitian Wang, Xinyu Yang, Minshen Zhang, Hao Zhang — Fri, 06 Nov 2026 00:00:00 +0000

97% · Efficiency 97%, New model 81%, Evaluation 45%

At context lengths of about 1M tokens, dense attention is costly in prefill and decoding, and training typically relies on long documents, which become scarcer as the target length increases. Most sparse attention methods score all preceding blocks or tokens, so their selection cost per query is $O(L)$ at context length $L$; hierarchical selection reduces it to $O(\log L)$, but prior trainable hierarchical methods report accuracy only within their training length. We propose RecSA (Recursive Sparse Attention), which builds a tree over the KV cache in which each node summarizes its subtree with one key and one mean-pooled value. The keys come from one pooler per layer, shared by all levels and distilled from dense attention. Each level above the tokens selects at most $k$ candidates, so each query has $O(\log L)$ selection cost and reads $O(\log L)$ entries; the summaries of unselected candidates share one softmax with the exact tokens. Each entry's RoPE position is its index among the candidates of its level, so at any length every relative distance stays within the range seen during training. Initialized from Qwen3-4B and trained on 0.62B tokens at the original model's 32K training length, RecSA outperforms all baselines on RULER at every length from 131K to 2M; at 1M, it scores 60.5, compared with 40.0 for the best baseline, Dense + DCA. With our optimized kernels, RecSA is faster than dense attention in end-to-end prefill and decoding from 64K to 2M; at 524K on one H200 GPU, it accelerates prefill by 3.2x and decoding by 3.8x.

★ Multimodal-SEA: Benchmarking AI Agent Security under Multimodal Social Engineering Attacks

Zhitao He, Chi Him Zhang, Chengru Han, Yi R Fung — Fri, 06 Nov 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 85%

Personal AI agents face social engineering attacks supported by forged multimodal content, but existing evaluations mainly focus on text-based dialogue, leaving their intrinsic security under such attacks insufficiently studied. We introduce Multimodal-SEA, an interactive benchmark comprising 608 expert-designed tasks grounded in real-world fraud cases, covering 14 domains and 152 attack scenarios with prepared multimodal content. The tasks span four difficulty levels and reflect the distinct social and cultural contexts of Chinese and English environments. Agents inspect content, verify requests, and execute actions in simulated applications, with attack outcomes grounded in verifiable tool execution logs. To enable controlled red teaming of personal agents, we design a multimodal attack harness that integrates task information, attack progression control, and multimodal content construction to evaluate their intrinsic security against adaptive social engineering attacks in simulated environments. Evaluating seven widely used multimodal models across six dimensions reveals that all are vulnerable, with attack success rates of 7.1% to 34.5%. Adding multimodal content increases attack success from 19.9% to 29.2%, despite increased proactive verification coverage. Stronger safety instructions reduce attack success without improving recognition of deceptive multimodal content, while lower attack success rates for some models accompany excessive refusal of legitimate requests.

★ StepBack: Step-Level Error-Localized Resampling for Efficient Test-Time Reasoning

Yang Ouyang, Jung-Eun Kim — Fri, 06 Nov 2026 00:00:00 +0000

97% · Efficiency 97%, Evaluation 39%

Self-consistency (SC) is a widely used decoding strategy for improving chain-of-thought reasoning, but it requires sampling many complete reasoning trajectories with a preset budget. Its efficient variants reduce the cost mainly by stopping full-trajectory sampling early or distributing the number of samples across questions. In this paper, we propose StepBack, a step-level error-localized resampling method for efficient test-time reasoning. Given a small set of draft trajectories, StepBack localizes likely first error positions using lightweight signals and then resamples the detected error step and its suffix, followed by majority answer voting. Empirically, StepBack matches the accuracy of SC on average across seven mathematical reasoning and multi-hop QA benchmarks, while reducing token usage by up to 41.26%. In addition, by nature, StepBack can be combined with adaptive inference methods such as Adaptive Consistency, further improving their inference-time efficiency.

★ Benchmarks as Software: Terminal-Bench and the Case for Continuous Benchmarks

E. Kelly Buchanan, Crystal Zhou, Scott W. Linderman, Christopher R\'e — Fri, 06 Nov 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 77%

Coding agents are among the most economically consequential deployments of LLMs to date, making reliable evaluation increasingly important. Yet even benchmarks that are carefully validated at release can become stale. As models improve, they can discover valid solutions that existing verifiers reject, while changes to evaluation infrastructure can invalidate assumptions embedded in tasks and tests. We call the resulting accumulation of defects benchmark maintenance debt: a growing gap between benchmark scores and the capabilities they are intended to measure. We introduce a continuous-validation workflow for agentic benchmarks that combines community reports, task artifacts, agent traces, automated oracle checks, and human-gated repairs. We apply this workflow to Terminal-Bench 2.0, a benchmark that underwent extensive human and LM-assisted review and red-teaming before release. Our audit identifies issues in roughly 30% of the benchmark, and we release the resulting repairs as Terminal-Bench 2.1. These repairs change evaluation results: accuracy increased by over 10 percentage points and leaderboard positions move by up to three ranks, even as the overall ranking remains highly correlated (Spearman rho = 0.958). We evaluate how resource allocation shapes benchmark conclusions. We show that fixed wall-clock can materially change agent performance: For example, extending the time limit closes a 20pp harness gap on GPT 5.4 and shrinks a 30 pp gap on GPT 5.4 mini to 5 pp. Benchmark maintenance is therefore not only about repairing task artifacts: resource budgets are part of the measurement specification and must be validated alongside the tasks themselves. More broadly, benchmark maintenance debt is not just accumulated implementation bugs: our results show that as models improve, both evaluation assumptions and resource settings may need to be revalidated.

★ ReproBridge: Building Reproducibility into ML Research Artifacts

Yubo Wang, Rui Meng, Jaehyun Nam, Tomas Pfister, Jinsung Yoon — Fri, 06 Nov 2026 00:00:00 +0000

97% · Evaluation 97%, Efficiency 81%, Agents 66%

Reproducing machine learning experiments requires reconstructing procedures from information scattered across papers, released code, and supporting artifacts. For coding agents, reproduction success can therefore depend on their ability to fill in missing details, rather than on the released artifacts alone. We ask whether preserving recovered procedures and decisions improves subsequent independent reproduction. We define a standardized artifact design comprising a refined codebase, environment setup, experiment launchers, and a reproduction guide, and introduce ReproBridge, an agent workflow for constructing these refined reproducible artifacts. ReproBridge-Bench provides 1,029 tasks from 387 experiments across 36 papers, anchored by five human-verified reference scores per task. Agents attempt tasks in fresh sessions with raw code or refined artifacts; a separate Auditor checks execution evidence. Across four evaluated models, refined artifacts raise mean reproduction success from 76.4% to 92.6% and reduce downstream API spending by 19.7% on average. The cross-model success-rate spread narrows from 21.6 to 3.9 percentage points, with all four models achieving mean success rates above 90%. Removing any component reduces success in the ablation study, with the largest losses from code and environment removal. These results suggest that preserving execution knowledge helps even models with lower raw-code success reproduce experiments more reliably and at lower API cost.

★ Are Frontier VLM Agents Ready to Be Robot Generalists? An Empirical Study with the Embodied Agent Arena

Haojian Huang, Pukun Zhao, Zexi Li, Yehang Zhang, Yangkai Wei, Wenqian Li — Fri, 06 Nov 2026 00:00:00 +0000

97% · Evaluation 97%

A robot generalist must interpret a scene, act on it, and satisfy every condition of a task. How far do frontier vision-language models (VLMs) meet these demands? We introduce Embodied Agent Arena, evaluating ten VLM agents across Geometry, Spatial Reasoning, Affordance, Task Planning, and Manipulation. Its 1,000 cases combine 32 established sources with GeoProbe, our 168-case benchmark for camera, motion, depth, and scale estimation. A minimal harness leaves task decisions to the model through source-native interfaces. Closed-weight models lead all five domains, whereas open-weight VLMs show useful spatial reasoning but substantially lower planning and manipulation success. Beneath these averages, similar geometric errors conceal different scale and shape distortions, and valid region estimates need not identify usable action points. During execution, agents can achieve early subgoals yet lose object control, bind an action to the wrong receiving region, or stop before the required physical state holds. Additional computation does not uniformly resolve these failures: code repair can restore execution without completing the task, while self-review can reverse correct judgments. Failure costs also differ: the strongest planners concentrate failed-token expenditure in prolonged body-control tasks, whereas other models fail broadly across procedures. Matched traces show how successful recovery uses feedback to restore a missing prerequisite and advance the remaining task.

★ TileQuant: Tile as the Precision-Allocation Unit for LLM Mixed-Precision Post-Training Quantization

Hanzhi Zhang, Qiao Zhang, Qinglei Cao, Heng Fan, Yan Huang, Kewei Sha — Fri, 06 Nov 2026 00:00:00 +0000

96% · Efficiency 96%

Post-training quantization (PTQ) reduces large language model (LLM) deployment costs, but preserving quality requires accounting for uneven weight sensitivity. Uniform PTQ uses one weight bit-width, importance-aware PTQ adjusts scales or reconstructs weights within that width, and column-wise mixed precision retains selected columns at higher precision. However, sensitivity varies within columns, so whole-column retention also protects less sensitive weights, while runtime outlier detection adds decoding overhead. We therefore introduce TileQuant, a training-free framework that assigns precision to hardware-aligned 2D weight tiles. Offline calibration provides input-activation-aware scoring from forward statistics for lower calibration cost and gradient-informed scoring from loss sensitivity for stronger quality preservation. Within each projection's retention budget, the highest-scoring tiles remain at 16-bit and the rest use 8-bit. The resulting map guides reconstruction into compact weight buffers, which the kernel reuses for native integer and 16-bit computation with local accumulation to reduce memory traffic. Experiments show that TileQuant reaches 2.66x the original model's generation throughput. Across language modeling, reasoning, long-context retrieval, and question answering, it preserves quality and surpasses quantized and original models on multiple tasks. These results support selecting precision units that capture local sensitivity while enabling efficient hardware execution. The implementation is: https://github.com/HanzhiZhang-Ulrica/TileQuant

★ EnvOpt: Automated Optimization for Agent Environments

Fri, 06 Nov 2026 00:00:00 +0000

96% · Efficiency 96%, Evaluation 46%, Agents 42%

An LLM agent's cost and success depend on its model, its harness, and its environment. In addition to the widely studied directions of parameter tuning and harness optimization, we propose the third direction: optimization over stationary environments. Agents are often deployed in rich environments shared by many tasks, such as a company's Slack workspace or a complex collection of data sources, documents or tickets. We call these environments stationary environments. We introduce EnvOpt, a method for automated end-to-end optimization of stationary environments for any given model and harness, using real or synthetic tasks without needing any graders or reference answers. On AutomationBench, EnvOpt reduces the cost of Claude Haiku 4.5 by 72.7% and raises its pass rate by 14.34%. The same environment transfers to Opus 5, cutting its cost by 68.1% with a 1.34% gain. Added on top of Meta-Harness's optimized harness for Haiku 4.5, it lowers cost by a further 83.4% and adds 11.17% accuracy over Meta-Harness alone. Across 5 benchmarks, EnvOpt significantly reduces the cost of agents while improving or maintaining pass rate, and is transferable to different models and different harnesses. We also discuss common successful and rejected environment optimization strategies during automated optimization. More broadly, our results show that how information is represented to an agent shapes both its cost and its success, making the environment worth optimizing in addition to the parameters and the harness.

★ SPARE: Fast Linear Attention Decoding with Compact Recurrent States

Ke Wan, Zaiyi Zheng, Chen Chen — Fri, 06 Nov 2026 00:00:00 +0000

96% · Efficiency 96%

Linear attention reduces the quadratic sequence-length complexity of softmax attention by maintaining fixed-size recurrent states. However, these states scale quadratically with the representation dimension and require repeated data transfers during decoding, creating a substantial GPU memory bottleneck. Recent studies have revealed significant low-rank redundancy in recurrent states, suggesting opportunities for more efficient representations. We introduce SPARE (State Projection for Accelerated Recurrent Execution), a training-free method that exploits this redundancy through fixed low-dimensional subspaces, heterogeneous rank allocation, and lightweight write compensation. By performing recurrent reads and updates directly on compact states, SPARE reduces memory traffic without online decomposition or dense-state reconstruction. SPARE achieves 2.55x full-model decoding speedup, 7.49x lower recurrent-kernel DRAM traffic, and 3.31x lower peak GPU memory, while preserving better associative-retrieval accuracy than structured pruning at matched memory budgets.

★ PharmAgents2: Towards Autonomous End-to-End Drug Discovery

Haichuan Tan, Bowen Gao, Tao Huang, Jiaxin Tan, Xin Hong, Baoguo Yao — Fri, 06 Nov 2026 00:00:00 +0000

96% · Evaluation 96%, Agents 85%

Drug discovery requires continuous scientific decision-making across a vast research space. We introduce PharmAgents2, an autonomous system for end-to-end drug discovery that combines knowledge-enhanced decision-making with organized large-scale exploration. Paper2Skill extracts transferable scientific decision knowledge from public sources and, together with expert-developed skills, supports research decisions. Self-Scaling Discovery expands exploration through parallel agents inheriting research context and decision-making capabilities across targets, scientific hypotheses, and discovery strategies. We introduce PharmAgents-Bench v0.1, comprising 121 questions across nine decision scenarios and three end-to-end tasks assessed with expert-defined process rubrics. PharmAgents2 outperforms evaluated baselines in most decision categories and achieves the highest process scores on all three end-to-end tasks. In three real-world studies, autonomous investigations generated findings subsequently supported by human-executed experiments. For LpxH, PharmAgents2 hypothesized a previously unreported allosteric site and screened compounds against it, identifying a novel inhibitor scaffold with experimentally confirmed inhibitory activity and binding. For JAK1, it prioritized a covalent inhibition strategy and nominated compounds with experimentally demonstrated activity. Starting from a known bioactive molecule, Self-Scaling Discovery explored candidate targets across the proteome and identified a new binding target, with the interaction subsequently confirmed experimentally. PharmAgents2 represents a prototype of a virtual pharmaceutical company, marking a step toward autonomous end-to-end drug discovery.

★ Continued Training of Continuous Language Flows with Token Identity Squared Supervision

Jiawei Jiang, Zhaoxiang Zang, Zhengtao Yu — Fri, 06 Nov 2026 00:00:00 +0000

96% · New model 96%, Evaluation 46%

Diffusion and flow models have recently been extended to language modeling, generating text through continuous denoising in token embedding spaces. Pretrained models provide token representations and contextual prediction capabilities, but improving generation with limited additional training requires an appropriate connection between supervision and continuous generation. Continuous MSE evaluates a predicted vector obtained by weighting token embeddings with predicted probabilities; different probability distributions can yield the same vector and are therefore indistinguishable under this objective. We propose Token Identity Brier Flow (TIBF), which applies squared supervision directly to token identities when continuing pretrained language flows. TIBF combines network scores with token-matching scores under a Gaussian observation model, supervises token probabilities using the Brier loss, and uses probability-weighted embeddings for continuous generation. With fixed token embeddings, we show that the squared error between predicted token probabilities and the true posterior bounds the squared errors of continuous denoising predictions and flow velocities. A risk decomposition separates uncertainty inherent in the observation from posterior fitting error, motivating continued training that preserves pretrained embeddings and focuses on probability prediction. We also establish convergence to token embeddings under ideal predictions and sufficient conditions for local convergence with prediction error. From common pretrained initializations and under matched continuation budgets, frozen-embedding TIBF achieves Gen.PPL values of 33.96 on OWT and 81.17 on LM1B, outperforming continuous-MSE and CE continuation controls. It attains the lowest Gen.PPL at every tested sampling budget on both datasets, together with the lowest repetition rate on LM1B.

★ CORE: Coarse-to-Fine Search with Context Management for Deep Research

Hoang Tran Vuong, Ngoc Quan Pham, Hung Bui, Fatih Porikli, Dat Quoc Nguyen — Fri, 06 Nov 2026 00:00:00 +0000

96% · Agents 96%, Evaluation 47%, Efficiency 31%

Deep research agents answer complex questions through repeated search. They rely on a harness that runs their tool calls and returns evidence to their context. This creates a trade-off: Search works best over long passages, which give enough context for reliable semantic matching. But the returned evidence should be short, because every observation stays in the agent's context for the rest of the trajectory. We introduce CORE, a training-free harness that combines coarse-to-fine search with context management. CORE retrieves passages with a dense retriever. It then reuses the token-level similarities from a late-interaction reranker to return only the most relevant snippet from each passage. Context management lets the agent expand the context around sources it has already retrieved and suppresses repeated snippets. In controlled closed-search comparisons with the Tongyi DeepResearch and OpenResearcher harnesses, CORE achieves state-of-the-art answer accuracy and evidence recall across three agent models and three benchmarks (BrowseComp-Plus, BrowseComp-Wikipedia and BrowseComp-Plus-CM), on corpora containing up to 553M documents.

★ Reinforcing What the Student Has Learned: Teacher-Guided On-Policy Distillation

Bozhen Fang, Linhao Zhong, Xuehao Wang, Huanyi Zheng, Hao Chen, Chunhua Shen — Fri, 06 Nov 2026 00:00:00 +0000

95% · New model 95%, Evaluation 34%

On-policy distillation (OPD) improves language-model reasoning through dense teacher feedback on student-generated responses.We first observe an amplitude disparity in this feedback: on math reasoning responses, style-related positions exhibit 3.5 times the mean absolute teacher--student log-probability difference of mathematical positions.Scaling OPD updates by these discrepancies thus directs much of the update strength toward matching the teacher's expression. We instead take the student's own pretrained base as a reference: the student--base difference records the token preferences acquired during post-training and provides supervision that does not depend on the student's distance from the teacher. Building on this reference, we propose Student-Delta On-Policy Distillation (SD-OPD), which uses centered student--base differences as update coefficients, keeps each coefficient only when its sign agrees with centered teacher--student feedback, and recomputes the student signal every training round. Across 5 teacher--student configurations with Qwen3 students of 1.7B to 8B parameters, SD-OPD improves six-task mathematics average pass@1 (avg@16) over OPD by 7.76 to 10.37 percentage points. Code is available at https://github.com/aim-uofa/sd-opd.

★ O$^2$PD: Overlap-Gated On-Policy Distillation on Reliable Prefix

Shuaishuai Zu, Jihao Zhao, Pengfei Li, Guanghuan Fang, Ben Feng, Yuhan Liu — Fri, 06 Nov 2026 00:00:00 +0000

95% · New model 95%

On-policy distillation (OPD) supplies dense token-level supervision by aligning a student to the teacher's next-token distribution on its own generated rollouts. Existing works mostly implement top-$k$ OPD, applying reverse Kullback-Leibler ($\RKL$) divergence between the teacher's and the student's renormalized distributions over teacher top-$k$ support. However, the teacher scores student-generated rollouts that are inherently off-policy for itself, leaving the reliability of its next-token guidance unclear. Restricting the divergence to the teacher top-$k$ support also discards the remaining mass outside that support. These gaps leave two questions implicit: 1) which tokens to align, and 2) how to align those tokens. To answer these questions, we propose Overlap-gated On-Policy Distillation (O$^{2}$PD). As for 1), we quantitatively analyze the reliability of teacher supervision and find that alignment on the student's trajectory prefix is more reliable than alignment on the full context. Therefore, our O$^{2}$PD is applied only to the 20\% prefix tokens of each student rollout. As for 2), we theoretically characterize two problems of top-$k$ OPD: the \emph{Teacher-support Uncovering} and the \emph{Dominant-Mode Mismatch}. Based on these analyses, we propose two support strategies: the \emph{Reset Support} strategy attaches a reset atom to every token, pooling mass outside the teacher top-$k$, and the \emph{Support Convergence} strategy converges the support to the already aligned intersection. Our O$^{2}$PD employs the designed overlap-gated $\alpha$-divergence to interpolate between mode-covering and mode-seeking for per-token alignment, where the overlap is used as a gate to switch the support and adaptively modulate $\alpha$.

★ PBPO: Prefix-Block Policy Optimization for Diffusion Language Models

Katherine Tieu, Dongqi Fu, Si Zhang, Hanqing Zeng, Tianxin Wei, Chenglin Wei — Fri, 06 Nov 2026 00:00:00 +0000

95% · New model 95%, Evaluation 36%

Diffusion Large Language Models (dLLMs) commonly decode responses by proceeding autoregressively across blocks while iteratively denoising multiple tokens within each block. This structure makes block-level policy optimization a natural adaptation of autoregressive policy optimization. However, the exact likelihood of a completed block is intractable because it marginalizes over the intermediate states and reveal orders of the denoising process. To realize the block-level formulation, we construct a prefix-block state and use a single model evaluation to define a factorized one-step surrogate for the block likelihood. The resulting block surrogate ratio is a product of its token surrogate ratios, allowing variation at individual positions to propagate across the entire block before clipping. We therefore introduce Prefix-Block Policy Optimization (PBPO), which preserves block-level conditioning but applies surrogate importance weighting and clipping separately to each token. PBPO further diversifies repeated updates by resampling corrupted views of the prompt and revealed prefix, a design motivated by masked-dLLM pre-training. Under our analysis assumptions, block-product weighting amplifies the variance of gradient contributions relative to token-local weighting. Across two LLaDA-family initializations, six benchmarks, and three generation lengths, PBPO achieves competitive performance against existing dLLM RL methods.

★ Do Recommenders Need to Think in Words? Latent Reasoning for Efficient LLM-Based Recommendation

Hongyang Su, Beibei Kong, Lei Cheng, Chengxiang Zhuo, Zang Li — Fri, 06 Nov 2026 00:00:00 +0000

95% · Efficiency 95%, New model 90%

LLM-based recommender systems increasingly generate chain-of-thought (CoT) text before each item prediction to activate the LLM's reasoning capacity, costing tens to hundreds of extra tokens per recommendation at inference. We ask whether inference-time reasoning artifacts (textual or latent) can be replaced by training-time supervision alone, without loss of accuracy. We propose LatentRec, a framework for LLM-based generative recommendation that replaces inference-time reasoning artifacts with training-time per-step head supervision. LatentRec has three components. A 2.5M-parameter Continuous Thought Module iterates K latent reasoning steps over the LLM's hidden state during training. Soft Token Selection bridges the ~14x norm gap between hidden states and input embeddings. A Per-step Reasoning Loss (PRL) supervises every intermediate state through the LLM's prediction head against the first token of the target Semantic ID. The trained LLM retains this accuracy without the module at inference, so the module can be discarded at deployment. On three Amazon benchmarks, LatentRec improves HR@10 by 4.0-5.0% (relative) over a CoT-distilled baseline at 18.3x lower latency (batch=1). Per-step training supervision consistently improves accuracy across three datasets, paid once at training rather than at every recommendation.

★ A Behavioural Framework for Predicting and Understanding Loss of Control in Frontier Artificial Intelligence Systems

Yao Huang, Yichi Zhang, Yifan Wang, Yitong Sun, Haoyu Dong, Jifan Ma — Fri, 06 Nov 2026 00:00:00 +0000

95% · Evaluation 95%, Agents 36%

Frontier artificial intelligence (AI) systems are increasingly operating over longer horizons with diminishing human oversight, raising the risk of loss of control: systems acting beyond effective human control with potentially irreversible consequences. Because such failures may become observable only after containment is difficult, they must be assessed prospectively. We introduce a behavioural framework that decomposes loss-of-control risk into three factors: misaligned motive, harm-enabling capability, and monitoring evasion. We operationalise these factors through 13 measurable aspects and 29 evaluation tasks, whose scores are combined into a single loss-of-control (LoC) score. Across 13 frontier models, the LoC score shows strong prospective validity, correlating with failure rates in contained long-horizon risk scenarios (Pearson r = 0.84). Risk varies systematically across models and cannot be explained by capability alone. We further show that the framework can guide targeted mitigations: fine-tuning and monitoring directed at diagnosed vulnerabilities reduce failure rates while largely preserving general performance. These results provide a practical basis for predicting, understanding, and mitigating loss-of-control risk before it materialises.

★ QRAE: Query-Relevant Attention Entropy for Context Compression in DeepSearch Agents

Yingjie Tan, Haiquan Qiu, Quanming Yao, Manyi Zhong, Xianzhi Yu, Yaqing Wang — Fri, 06 Nov 2026 00:00:00 +0000

95% · Efficiency 95%, Agents 94%, Evaluation 46%

DeepSearch LLM agents repeatedly search and carry retrieved sources into subsequent reasoning steps. This expanding context increases token consumption and latency, although many results do not contribute to later reasoning. We propose Query-Relevant Attention Entropy (QRAE), an online context compression method that uses sparse evidence-assessment heads to decide which intact retrieved chunks should persist across reasoning steps. It scores bounded candidate batches through paired prefills: a query-free control suppresses shared attention responses, while normalized entropy measures the breadth of the positive response within each chunk. High-scoring chunks enter a persistent evidence whitelist and are included in the reasoning context. Uncertain chunks remain outside the reasoning context but are retained in a candidate cache, where they can be retrieved for later reevaluation. QRAE requires neither an additional generative compression pass nor modification of the attention operator. Across multiple DeepSearch benchmarks and reasoning models, QRAE improves task success by an average of 2.35 percentage points and reduces end-to-end latency and reported token usage by an average of 30% and 45%, respectively. These benefits generalize to a model with nearly 100B parameters.

★ Agent-R$^2$: Scaling Agents for Realistic Productivity Work via Real-Source Data Synthesis

Teng Pan, Zixuan Wang, Yuchen Yan, Qifei Wu, Yao Chen, Yingqi Qu — Fri, 06 Nov 2026 00:00:00 +0000

95% · New model 95%, Agents 93%, Evaluation 71%

Large language models have made rapid progress in agentic reasoning and tool use, raising expectations for agents that can take over realistic productivity work from humans. However, current agents still struggle to finish such tasks, where success requires following long-horizon professional workflows, understanding heterogeneous file contexts, and producing usable deliverables under fine-grained requirements. A key bottleneck is the lack of scalable post-training data that reflects real work intent, authentic file environments, and deliverable-level evaluation. To close this gap, we introduce Agent-R$^2$, a $\textbf{r}$eal-source data synthesis framework that scales training data for $\textbf{r}$ealistic productivity work. Instead of relying on manually specified workflows or LLM-generated files, Agent-R$^2$ grounds synthesis throughout the pipeline in real occupational seeds, corpus-derived workflow evidence, and authentic web-sourced files. From 10k web-corpus candidates, Agent-R$^2$ distills 4k task instances that pair real multi-format attachments with fine-grained, deliverable-level rubrics. On this data, we train Agent-R$^2$-4B via supervised fine-tuning and rubric-based reinforcement learning, improving both end-to-end productivity performance and file-grounded reasoning. Trained on only 4k instances, Agent-R$^2$-4B surpasses Qwen3-235B-A22B-Thinking-2507 on both GDPval (51.0$\%$ vs. 48.0$\%$) and OfficeQA (14.8$\%$ vs. 13.8$\%$). Further analysis shows that Agent-R$^2$ mitigates shortcut behavior, where agents produce superficially plausible deliverables before sufficiently grounding, constructing, or verifying them, while promoting sustained exploration.

★ SpecCTMC: Exact Speculative Simulation of Diffusion Language Model Continuous-Time Markov Chains

Siva Rajesh Kasa, Yasong Dai, Sumit Negi, Hongdong Li — Fri, 06 Nov 2026 00:00:00 +0000

95% · Efficiency 95%, Evaluation 31%

Masked diffusion language models offer parallel text generation, but independently sampling multiple tokens can disrupt their dependencies and change the output distribution. Exact continuous-time sampling avoids this approximation, yet typically requires one sequential model evaluation per token. We introduce SpecCTMC, a training-free speculative sampler that reduces this sequential bottleneck without a separate draft model. SpecCTMC proposes successive token updates using previously computed model predictions, evaluates their intermediate states in parallel, and applies an exact acceptance-and-correction procedure. We prove that SpecCTMC preserves the joint distribution of event times, generation order, and token values of the underlying continuous-time Markov chain. It therefore introduces neither time-discretization error nor additional factorization error from parallel decoding. Experiments on LLaDA-8B-Instruct and Dream-v0-Instruct-7B across five benchmarks demonstrate 1.73-4.38x fewer sequential model evaluations than direct continuous-time sampling, while preserving the reference sampling distribution. Our code is available at: https://github.com/kasakh/specctmc

★ Muir: Linear Attention with Sensitivity-Allocated Gating and Matched-Filter Readout

Tushaar Gangavarapu, Gil Tabak, Clemens JS Schaefer — Fri, 06 Nov 2026 00:00:00 +0000

95% · New model 95%, Efficiency 62%

Linear attention compresses context into a fixed-size key-value state for linear-time training and constant-memory inference. Gated delta-rule models decay this state and erase its prediction for an incoming key before writing a new value; Gated DeltaNet-2 (GDN-2) generalizes this with channel-wise decay and decoupled, full-rank erase and write gates. However, we show that state updates are far less sensitive to erasure than to writes: key activation and unit normalization concentrate key energy onto a sparse subset of channels, whereas value writes scale all channels, so the erase gate can be parameterized at lower rank. Furthermore, slow-decaying state rows accumulate much larger norms than fast rows, dominating unweighted readout and forcing stronger write-time erasure to suppress interference. Guided by this analysis, we introduce Muir, which fuses all gates into a sensitivity-allocated low-rank projection, re-weights state rows with a matched-filter readout gate, and reinvests the saved parameters in a wider feed-forward network. Because standard chunkwise algorithms become numerically unstable at the large chunk sizes needed to saturate wide matrix units, we design a parallel update for stable, efficient training. Across 166M-1.34B parameters, Muir lowers perplexity by 1.8-2.0% over GDN-2 and 2.5-3.6% over Transformer, raises commonsense accuracy by up to 1.9 points, and improves extractive recall over GDN-2. At 1.34B, Muir outperforms both baselines on multi-hop and multi-needle recall, extrapolates 2,560x from its 4K training length to 10M-token streams without perplexity drift, and maintains near-100% passkey retrieval to 1M tokens with static readout. Muir achieves these gains while matching GDN-2 throughput, reaching its 100B-token validation loss 1.2x faster, and training 2.7x faster than Transformer at 32K.

$D^2$-OPD: Separating Direction and Distance in Extrapolative On-Policy Distillation

Fri, 06 Nov 2026 00:00:00 +0000

94% · New model 94%, Evaluation 38%

On-policy distillation (OPD) trains a student model to match a teacher's next-token distributions on student-generated prefixes. Extrapolative OPD further extends the teacher's log-probability shift relative to a reference policy to construct stronger learning targets. However, existing methods couple extrapolation direction and distance, allowing the magnitude of the teacher's shift to dictate how far the target moves. We propose $D^2$-OPD, decoupling these two factors: a per-prefix hard KL constraint bounds the target's distance from the teacher, while an adaptive combination of the teacher's post-training shift and the initial student--teacher gap determines the extrapolation direction. We optimize the student toward this target using forward KL, providing learning signals even for low-probability tokens. We evaluate three settings: where standard OPD performs well, where it struggles with base-model initialization, and where no teacher reference is available. On mathematical reasoning benchmarks, $D^2$-OPD attains the highest score in all three settings, outperforming the strongest baseline by up to 9.45 points. Ablations validate the contributions of each component.

HyFlash: A Prefix-Memory Prior for On-Device Speculative Decoding

Jongwoo Ko, Pashmina Cameron — Fri, 06 Nov 2026 00:00:00 +0000

94% · Efficiency 94%, New model 67%

Parallel block drafting accelerates speculative decoding by predicting multiple future positions in a single backbone pass, but on-device deployment constrains both drafter capacity and context storage. Under tight memory budgets, drafters must model long-range dependencies and fine-grained draft-position interactions without the growing KV-cache overhead of full-context attention. We propose HyFlash, a compact parallel drafter that introduces a prefix-memory prior to separate long-range context modeling from fine-grained draft-position interactions with limited capacity. Local attention captures recent context and draft-position interactions, while recurrent memory compresses historical target features into fixed-size states. This inductive bias enables effective local and long-range modeling with bounded draft-side memory. Experiments show that HyFlash achieves higher acceptance than full-context attention with substantially less KV storage and with larger gains for shallow drafters and long contexts.

SAVeR$^2$: Reasoning-based Safety Alignment for Large Reasoning Models via Verifiable Rewards

Yichen Sun, Qirui Liu, Yan Wang, Jianan Lin, Linbo Jiang, Shiyu Wang — Fri, 06 Nov 2026 00:00:00 +0000

94% · New model 94%, Evaluation 44%

Large Reasoning Models (LRMs) have achieved impressive performance via explicit Chain-of-Thought (CoT) reasoning, yet this process introduces critical safety risks. We formalize LRM safety objectives as requiring non-malicious reasoning and answers, safety consistency, and explicit intent identification. We argue that existing alignment methods suffer from poor adversarial generalization, a significant ``safety tax'' on reasoning, and unfaithful coarse-grained rewards, primarily due to their neglect of the LRM's intrinsic reasoning capabilities for safety. To address these, we propose \textbf{SAVeR$^2$}, a novel \textbf{S}afety \textbf{A}lignment method via \textbf{R}ewards with \textbf{R}easoning capability. SAVeR$^2$ decomposes the safety objective into four fine-grained verifiable reward signals and utilizes a difficulty-aware data selection strategy to stabilize training. Crucially, we introduce a reasoning-preserving gradient projection mechanism that analytically resolves conflicts between safety and reasoning gradients, effectively eliminating the safety tax. Extensive experiments on models ranging from 1.5B to 671B demonstrate that SAVeR$^2$ achieves superior safety performance and significantly lower over-refusal rates across multiple benchmarks without compromising general reasoning capabilities.

Enhancing Repository-Level Code Completion with Reinforcement Learning in Real-World Scenarios

Zhenlong Dai, Zhuoluo Zhao, Xiu Tang, Sai Wu, Chang Yao, Zhipeng Gao — Fri, 06 Nov 2026 00:00:00 +0000

94% · New model 94%, Evaluation 43%

The rapid development of large language models (LLMs) for code, which closely aligns with the practical needs of real-world software development, has sparked growing attention in repository-level code completion. However, existing Code LLMs present challenges in focusing on suitable contexts from the repository and in deeply reasoning about cross-file dependencies. To address these challenges, we propose a novel reinforcement learning framework for repository-level code completion. To better understand both in-file and cross-file contexts, we employ identifier-driven intent recognition to capture completion intent, thereby improving the model's performance in real-world scenarios. To enhance cross-file reasoning, we propose a reward-driven completion learning to effectively reward in complex repository completion scenarios. To guide LLMs in completing code that aligns with intended functionality and repository context, we introduce a selective-exploration strategy that directs the model to focus on low-confidence, high-reward tokens, promoting the exploration of valuable, underexplored completion patterns. Experimental results show that our approach significantly improves the performance of Code LLMs. The code and the dataset are available at https://github.com/zhenlongDai/RepoCompletion.

Persistent Retrieval for Conversational Agents: A Study of Retrieval Efficiency

Karthik Ravula, Sathwik Kowtha, Rishabh Verma, Kala Mehak Jain, Gunjit Mittal — Fri, 06 Nov 2026 00:00:00 +0000

94% · Efficiency 94%, Agents 78%, Evaluation 60%

Conversational agents can retrieve information directly or delegate retrieval to a subagent. This design choice affects how retrieval history is maintained, what context each model call receives, and what information must be re-retrieved in follow-up questions. We compare three designs: skill-based retrieval in the main agent, a non-persistent retrieval subagent that generates a findings response, and our persistent retrieval subagent, which maintains retrieval history across turns and returns retrieved evidence directly to the main agent. Our approach also allows the main agent to request corrections from the same subagent. We evaluate the three designs on enterprise questions posed independently, within shared conversations, and as dependent follow-ups. Across five multi-turn settings, our approach uses 43.3% fewer recorded agent-model tokens and incurs 45.0% lower estimated API cost than skill-based retrieval, while reducing average response time by 9.1%. Compared with non-persistent retrieval, it reduces response time by 15.6 seconds per question, while using 65.1% more tokens and incurring 29.1% higher estimated API cost. Our approach shifts the retrieval subagent from producing synthesized responses to maintaining retrieval state and supplying evidence that the main agent can directly evaluate and use for answer generation.

ChemE-as-Code: An Open-Source Platform for Agentic AI in Chemical Engineering

Artur M. Schweidtmann — Fri, 06 Nov 2026 00:00:00 +0000

94% · Agents 94%, Evaluation 44%, Hallucination 41%

ChemE-as-Code provides an open-source environment in which AI agents built on large language models (LLMs) can carry out chemical engineering work and engineers can check it. Its calculation tools range from catalysis through process simulation to plant layout, HAZOP and life cycle assessment, and skills tell a general-purpose agent how to use them. Engineering files are linked, machine-readable and versioned in Git; calculations run outside the LLM, each value records its origin, and automated checks report inconsistencies and out-of-date results by file and line. In an illustrative ammonia case, agents developed a written task into a preliminary basic-engineering package (simulation files, P&IDs, layout, HAZOP, datasheets, relief) for USD 224 of API cost in the design sessions. Planted faults and engineering changes show that the checks find what their rules cover and miss the rest, so they guide human review rather than replace it. No independent engineer has reviewed the design.

CoRD: Coupled Rollouts Fill the Blind Spot of On-Policy Distillation

Guhan Chen, Bohan Li, Jun Rao, Hejin Wang, Songtao Tian, Yexin Xie — Fri, 06 Nov 2026 00:00:00 +0000

94% · New model 94%, Evaluation 37%

On-policy distillation trains a student on its own rollouts, so states the teacher would reach but the student does not are seldom supervised. We introduce CoRD (coupled-rollout distillation), which keeps the on-policy loss and moves only the distribution of training states toward the teacher, drawing rollouts from a geometric interpolation of student and teacher through the maximal coupling of speculative decoding. The coupling keeps the student's token wherever the two distributions agree and otherwise substitutes a token from the part of the target the student underweights, so interventions are as rare as any exact sampler allows and recede as the student learns. Across six base students from two model families (Qwen3 and Qwen3.5, 0.6B to 4B), CoRD raises average pass@1 over on-policy distillation by 0.9 to 2.0 points and pass@8 by 1.2 to 3.4 points on five public math benchmarks at equal steps and data. On frozen teacher-reached states, a 4B student trained with CoRD reaches in 16 steps a gap closure that on-policy distillation does not reach in 80, and substitution rules intervening at least as often in other directions fall short of CoRD.

Random Is Almost Enough? Stable and Low-Cost Benchmark Compression via BestRand

George Ling, Shanshan Zhong, Lijin Yang, Hao Yang, Nanxiang Li, Zhongzhan Huang — Fri, 06 Nov 2026 00:00:00 +0000

94% · Evaluation 94%, Efficiency 86%

Benchmark compression reduces the often prohibitive time and cost of model evaluation by evaluating models on a small benchmark subset that closely matches full-benchmark results. To quantify estimation error, we derive an accurate analytical approximation to the error distribution when estimating full-benchmark scores from a subset, showing that most subsets yield small errors, while a few produce much larger errors, resulting in a long-tailed distribution. This means random sampling alone can achieve surprisingly strong performance with high probability, but remains unstable due to rare large errors. We further show that this instability is inherent to modern foundation models and persists even as more models are evaluated. Based on these findings, we introduce BestRand, a stable and low-cost benchmark compression method that preserves the efficiency of random sampling while substantially reducing its high-error tail. We validate BestRand on 70+ text, multimodal, and agent benchmarks, and further extend it to VLA evaluation. Compared with competitive methods, BestRand reduces high-error risk by 50.2% and estimation error by 16.4% on average, while requiring only 51.3% of the construction cost. It further reduces OOD estimation error by 21.8% on average across score-range, model-family, and temporal shifts. Together, our theory provides practical guidance for understanding and predicting the accuracy and reliability of benchmark compression, while BestRand enables stable and accurate compression at low cost. The code and data are publicly available onhttps://github.com/George-Ling3/BestRand.

SAFE-Teen: Evaluating Chatbot Safety Across Adolescent Mental Health Conversations

Fri, 06 Nov 2026 00:00:00 +0000

93% · Evaluation 93%

Adolescents are increasingly turning to AI chatbots for mental health support, advice, and companionship, raising safety concerns. While recent evaluations have started to examine safety across multiple conversational turns, few compare how slight changes in a teen's responses can affect a chatbot's safety in the same scenario. We conduct a clinician-informed evaluation of chatbot safety in simulated conversations with adolescents aged 13-17, examining how safety failures vary with adolescents' responses and teen-aware system prompting. We evaluated 32,555 simulated conversations across 18 chatbot APIs and two system-prompt conditions using 22 safety criteria grounded in previously identified harms. Conversations spanned 983 variants of 106 seed scenarios covering urgent safety issues, health advice, and identity and relationships. The overall criterion failure rate was 43.3% without the teen-aware prompt and 24.9% with it. The teen-aware prompt reduced failures, but it didn't eliminate them. When a teenager pushed back on a chatbot's answer, failures on mental health diagnosis, medication, and substance misuse boundaries rose by 10-12 percentage points without the prompt, for more than 80% of chatbots. These findings reveal persistent safety failures in simulated adolescent-chatbot interactions and show that safety depends on how a conversation unfolds.

Who Should Know What? Empirical Information Design for Multi-Agent LLM Economies

Mete Erdogan, Mirac Suzgun, Aneesh Pappu, Guowei Xu, Batu El, Andrew A. Schoen — Fri, 06 Nov 2026 00:00:00 +0000

93% · Evaluation 93%

The performance of multi-agent LLM systems depends not only on the capabilities of the underlying models but also on information-sharing policies, which govern what information each agent observes and at what level of detail. But does providing AI agents with more detailed information always improve collective outcomes, or can transparency paradoxically degrade efficiency? To investigate this, we introduce empirical information design, an experimental framework that systematically compares disclosure policies across economic markets populated by LLM agents, while holding interaction rules and models fixed. We evaluate ten frontier LLMs in multi-resource trade, bilateral bargaining, and double auctions, using ground-truth utilities to decouple identified opportunities from realized outcomes. We find that finer information disclosure yields highly model-dependent results. In trade, revealing exact valuations rather than ordinal preference rankings reduces realized gains for GPT-6 Sol (from 98% to 91% of maximum) yet increases them for Claude Sonnet 5 (from 72% to 86%). In bargaining, exact information leads Grok 4.7 to produce less balanced agreements, whereas Claude Opus 5.5 divides gains symmetrically. In double auctions, full disclosure inflates price demands, triggering market breakdown and lowering welfare for Gemini 3.1 Pro (97% to 78%), Claude Opus 5.5 (94% to 64%), and Grok 4.7 (97% to 62%). Overall, these results suggest that model selection and information design must be co-optimized, as LLMs vary in how their strategic responses to information impact system objectives.

RIVER: Evidence-guided multi-agent retrieval for enzyme--reaction discovery

Yifan Chang, Yunze Li, Feifei Jiang, Juntao Wu, Xin Chen, Han Hu — Fri, 06 Nov 2026 00:00:00 +0000

93% · Agents 93%, Evaluation 61%, Hallucination 41%

Enzyme function annotation and the identification of candidate enzymes for a given reaction are essential steps in connecting genomic information to biochemical reaction networks. However, existing methods rely on sequence or structural similarity or on training with known enzyme--reaction pairs, making it difficult to integrate biochemical evidence across large candidate spaces; they also struggle with retrieval tasks involving enzyme families whose members are highly similar. Here, we introduce RIVER (Reaction--enzyme Inference via Evidence-guided Retrieval), a multi-agent framework for bidirectional enzyme--reaction retrieval that identifies potentially catalyzed reactions for a given enzyme and ranks candidate enzymes for a given reaction. RIVER coordinates large language models, specialized biochemical computational tools, and external knowledge resources to assemble evidence at multiple levels, including reaction chemistry, protein sequences and domains, catalytic mechanisms, and database annotations. Dedicated validation modules further examine evidence gaps and potential conflicts and rerank candidates using traceable evidence records. On public benchmarks, RIVER achieves the highest success rates in both enzyme retrieval and enzyme annotation. In family-specific enzyme retrieval, RIVER outperforms the evaluated baselines on the selected datasets and metrics. In Escherichia coli enzyme retrieval tasks constructed from recent literature, RIVER achieves the best retrieval results with closed-source and open-source model configurations, substantially outperforming closed-source general-purpose and biology-specialized multi-agent systems, including Claude Code, Codex, and Phylo Max. These results suggest that evidence-guided multi-agent retrieval can prioritize biochemically supported enzyme--reaction relationships, providing a basis for subsequent experimental validation.

Towards Foundation Decision Models: Drex

Fri, 06 Nov 2026 00:00:00 +0000

93% · New model 93%, Efficiency 90%, Evaluation 42%

Autoregressive approaches with large language models, such as long chain-of-thought reasoning, have become the default for new AI applications because they reach higher accuracy on a wide range of tasks. In production, however, they give inconsistent answers, require parsing of outputs, and add cost and latency. Decision models address this: given a question and a set of options supplied by the user, they return a probability distribution over the options without autoregressive decoding. We present Drex, a family of four open decision models with 8B (DLM), 9B (v1.5), 35B with 3B active (v2.0), and 27B (Max) parameters. Each request consists of a state (a document, record, or agent context, optionally with images) and questions about it, each asking for a choice among options, a yes/no judgment, or a score on an ordered scale. Training combines supervised fine-tuning with reinforcement learning from adjusted feedback (RLAF), plus vision tuning for Max. Ablations show that RLAF improves accuracy and that consistency losses improve robustness to option order. On Decision Index 0.4.1, Max sets a new state of the art among the decision models we compare (67.52 on the public suite vs. 60.70 for Jev 1.14.0). On Decision Bench v4, it is statistically tied with d3-turbo and GPT-6 Sol and trails Claude Opus 5.5 by 1.6 points, and it has the highest accuracy among decision models on DecideBench. Max answers a typical request in 183 ms, against about 2.0 s for GPT-6 Sol and 3.5 s for Claude Opus 5.5, and its outputs are stable under option reordering. All four models are open source: https://github.com/nace-ai/drex-decision-models

Beyond K/V Caches: Cross-Model State Handoff in Hybrid Linear-Attention LLMs

Jiacheng Cai, Tao Zhang, Siqi Liang, Yun Dai, Ziqian Zeng — Fri, 06 Nov 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 56%

Production deployments often use different models to balance cost and quality through cascading, routing, and switching models during long conversations. Context accumulates across turns. Each switch requires the target to repeat prefill over this context. Existing methods transfer full-attention (FA) K/V caches across models. Hybrid linear-attention models also retain recurrent states (REC) and convolution histories (Conv). The interaction between Conv and REC complicates cross-model state transfer. Conv perturbations can change keys and values in recurrent layers and alter REC updates. Conv histories are overwritten, but changes in REC can still affect later outputs. More accurate REC reconstruction need not improve continuation. Building on this, we propose Hybrid State Transfer (HST) for all three state types. It uses structured REC maps and preserves history order during Conv alignment. We calibrate the adapters using the frozen target's outputs during continuation. FA cache map updates are merged offline. Across five benchmarks, Qwen3.5 4B$\rightarrow$27B and 9B$\rightarrow$27B achieve mean task retentions of 89.88% and 94.11%, respectively. With source caches already available, separate timing on synthetic 16K inputs yields handoff body speedups of $21.66\times$ and $21.91\times$.

Measured Coupling for Parallel Decoding in Diffusion Language Models

Tal Oved, Lior Wolf — Fri, 06 Nov 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 60%

Diffusion language models can write many tokens per step, but must choose which tokens are safe to commit together. Mean-field decoding penalizes pairs of positions whose predictions look alike. We replace this proxy with the coupling itself: how much committing one token changes the model's confidence at another. Cheap probes inside the main forward pass measure it, with no training or extra model, and it plugs into the same mean-field rule. On three models and four benchmarks, our method needs far fewer steps than mean-field decoding and is faster in most settings, also with a key-value cache, at an average accuracy cost of 1.1 points. Shuffling the couplings or replacing them with a uniform bonus loses this gain.

Scientist World: Grow a Research World for AI-Scientists

Taiming Lu, Sachin Konan, Zhuang Liu — Fri, 06 Nov 2026 00:00:00 +0000

93% · Evaluation 93%, Agents 54%

AI-Scientists arrived, but a world for them to grow in did not. Human scientists take theirs for granted $-$ centuries of open problems and shared ideas, waiting in the scientific literature. To AI-Scientists, that world is readable but not runnable, so we build Scientist World for them: every research paper becomes an executable environment. Our autonomous pipeline extracts each paper's open-ended research question and reproduces its proposed method as the baseline; an AI-Scientist entering the environment is evaluated against what the human authors achieved. We release the first 1,000 verified environments. With the world as an evaluation, we found that frontier AI-Scientists can already surpass the human researchers 74% of the time. As a record of how AI does science, it contains 6,400 hours of research traces, with every idea, experiment, failure, and success. As a training ground, we are able to raise (+158%) a small model to match frontier models in research ideation. At last, AI-Scientists' own world has arrived, grown from and growing with science itself.

Octopus: One Brain, Many Branches for Latent-KV Multi-Agent Reinforcement Learning

Tianming Sha — Fri, 06 Nov 2026 00:00:00 +0000

92% · Agents 92%, Efficiency 91%, New model 85%

Multi-agent LLM systems explore in parallel by spawning sub-agents with isolated context windows, whereas single-agent systems append every search and verification step into one growing trajectory. However, existing multi-agent frameworks duplicate prefix KV caches across branches, exchange findings by re-prefilling text summaries, and flatten or drop cross-branch dependencies during reinforcement learning (RL). We introduce Octopus, a single-model multi-agent RL framework in which all threads share one backbone and coordinate over a GPU-resident directed acyclic graph (DAG) of KV states through five primitives (fork, read_thread, wait_threads, rewind, commit). Octopus combines three co-designed mechanisms: (1) Zero-Copy Radix Forking, which spawns branches as $\mathcal{O}(1)$ cursor views over shared physical KV pages; (2) Attention-Only KV Mounting, which mounts a sibling thread's terminal KV blocks directly into full-attention layers while bypassing linear-recurrent states (Gated DeltaNet / Mamba-2), cutting off-task hijacking by $21\times$ ($57.9\% \to 2.7\%$) while retaining $92.2\%$ of full-text mutual information at zero extra KV allocation; and (3) a Packed DAG-GRPO Trainer, which packs all episode branches into one block-sparse attention pass and assigns counterfactual branch credit via leave-one-branch-out marginal rewards. Trained end-to-end with GRPO on Qwen3.5-4B over a sparse BM25 corpus, Octopus reaches $49.6\%$ accuracy on BrowseComp-Plus ($+17.5$ points over single-agent GRPO at 143K context and $+12.5$ points over text-summary multi-agent RL) and $51.2\%$ on OpenResearcher, outperforming dense-retrieval agents up to $250\times$ larger while increasing training throughput $3.3\times$ and cutting peak KV memory $2.8\times$.

The Last Mile to the Real World: Agent-Native Primitives as Reusable Building Blocks for Solving Real-World Tasks

Haibo Jin, Peng Kuang, Xiaopeng Yuan, Lucheng Fu, Jerry Wang, Xucheng Yu — Fri, 06 Nov 2026 00:00:00 +0000

92% · Agents 92%, Efficiency 85%, Evaluation 73%

How can an agent handle a real-world task end to end? Frontier models with their own harnesses still fall short on real-world benchmarks: on Terminal-Bench 4.0, GPT-5.6 Sol with Codex resolves 37.3% of the tasks and Claude Opus 5 with Claude Code 51.8%. These tasks fail at the step from reaching a resource to applying it, which we call the last mile to the real world. Existing approaches put the resource within reach but leave the competence to apply it with the caller, so each harness rebuilds that competence. We introduce agent-native primitives, reusable units that equip a resource with a resident LLM that applies it in context and communicates with other primitives, and HEELO (Harness Engineering around Evolving Agent-Native Primitives for the Last Mile to Real-WOrld Tasks), a harness engineering framework that builds each task's harness from a library of such primitives. HEELO composes the primitives a task needs into a task-specific graph, verifies the result against evidence from the environment, revises the primitives a failure implicates, and admits new or improved primitives after validation. We instantiate HEELO with eight primitive families, each built for one kind of task, and evaluate them on more than twenty benchmarks. Dev-Primitives raise Terminal-Bench 4.0 resolution from 37.3% under Codex to 55.5% on the same backbone, Harness Primitives composed per task raise Terminal-Bench 2.0 Pass@1 from 60.0% for the seed harness to 72.2%, Tool Primitives raise ToolBench pass rate from 55.8% to 75.1%, and Navigation Primitives raise GAIA-Text-103 accuracy from 45.6% to 58.3%. On the 46 real-world tasks of our heelo.dev platform, Claude Opus 5.5 with Claude Code and GPT-6.1 Sol with Codex complete 9 each on their own and 39 and 35 when connected to HEELO through our MCP server, with 1.9$\times$ and 2.3$\times$ less time and 1.55$\times$ lower cost per completed task.

FlashOffloading: Low-Latency KV Cache Offloading for Long-Context LLM Serving

Xuan Wang, Zhuanhao Wu, Tingzhu Fang, Xushi Li, Sicheng Zhang, Zhenyu Yang — Fri, 06 Nov 2026 00:00:00 +0000

92% · Efficiency 92%, Evaluation 31%

In large language model (LLM) serving systems, the key-value (KV) cache grows linearly with sequence length and concurrency, so long-context inference easily hits the high-bandwidth memory (HBM) capacity wall and results in low concurrency. Offloading the KV cache to host dynamic random-access memory (DRAM) frees HBM space and raises concurrency, but the additional cache management and host-to-device (H2D) load overhead inflates inference latency, higher concurrency fails to translate into effective throughput. We present FlashOffloading, a host-DRAM-centric inference architecture that fully offloads the KV cache in both the prefill and decode phases. FlashOffloading builds rolling HBM cache for prefill, a hot cache for decode, a reverse sparse index with zero extra miss-lookup overhead, and a compute--load overlapped kernel, optimizing KV offloading latency across the entire pipeline and delivering throughput gains under a time-per-output-token (TPOT) service-level objective (SLO). In experiments deploying GLM-5.2 with multi-token prediction (MTP3) speculative decoding on an Ascend910C prefill--decode (PD)-disaggregated platform, FlashOffloading improves prefill throughput by up to 1.38$\times$ and decode throughput by up to 1.91$\times$ under a 30 ms TPOT constraint. FlashOffloading is integrated into an open-source serving framework: https://github.com/vllm-project/vllm-ascend/tree/releases/v0.30.0.

Stress Testing Chain-of-Thought Monitoring Against Covert Misalignment

Yao Huang, Yifan Wang, Yitong Sun, Yichi Zhang, Yinpeng Dong — Fri, 06 Nov 2026 00:00:00 +0000

92% · Evaluation 92%

As LLM agents advance rapidly in autonomy, the risk of hidden objectives that bypass safety constraints becomes increasingly plausible. Chain-of-thought (CoT) monitoring offers a key control mechanism by reading the agent's reasoning and flagging misalignment before it materializes into action, yet its reliability depends on the realism of the red-team it is stress-tested against. Existing red-teaming efforts rely on shared templates, such as casting a side task as a prerequisite, which treat the side objective as an exogenous insertion and leave task-specific, contextually grounded evasions underexplored. We therefore ask: How robust is CoT monitoring against agents that strategically conceal misalignment? To answer this, we introduce TraceMRT, an automated Trace-level Monitor Red-Teaming framework that reframes the side task as a coherent sub-goal of the primary task and searches trace-shaping strategies whose induced trajectories complete the misaligned objective while reading as natural execution. Through black-box attacks on state-of-the-art monitors such as GPT-5.2 and Claude-4.5, we reveal substantial vulnerabilities even in top models. To counter these vulnerabilities, we propose a Bayesian Trajectory Monitoring framework, integrating global context and local segment evidence to detect covert misalignment. Empirically, our method significantly consolidates monitor robustness with 0.91 AUC and 0.74 TPR@FPR=0.05.

Don't Prompt Your Agents. Equip Them: Skillware as a Modular Standard for Sovereign, Auditable, and Deterministic Capability Execution for Logical Systems

Vladimiros Peilivanidis (ARPA Hellenic Logical Systems) — Fri, 06 Nov 2026 00:00:00 +0000

92% · Efficiency 92%, Agents 86%, Evaluation 86%

Current autonomous logical systems rely on a fragile assumption: that language models should dynamically write code, guess API payloads, and memorize operational constraints through prompt engineering. In practice, this burns massive context windows, inflates API bills, and produces brittle agents that fail unpredictably in production. In this work, we present Skillware, an open architectural framework and capability registry that formally separates what an agent wants to do from how that task is executed. Instead of prompting models with loose text instructions or markdown files baptized as 'skills', Skillware equips them with modular, versioned software capability bundles. Much like skill hotkeys in video games that execute deterministically under fixed world rules, each bundle combines a typed JSON input contract, deterministic Python code, task directives, bundled datasets, and automated unit tests. In-process adapters translate these capabilities into native tool declarations for any model, from frontier cloud APIs to local open-weight runtimes with zero network latency. Evaluating 23 frontier and open-weight models across 368 benchmark conditions, we show that small local models (3B to 8B parameters) equipped with Skillware consistently outperform unequipped frontier models on complex multi-step workflows. They cut token usage by up to 84.8%, slash inference costs by up to 97.3%, and eliminate parameter errors entirely. True agent reliability does not come from larger models or longer prompts, but from equipping logical systems with verified, deterministic software. Finally, we articulate the trajectory of capability bundles across an invariant continuum of information processing systems: from digital logical systems today, to IoT, drones, and embodied robotics tomorrow, establishing a foundation for auditable, real-time capability execution.

Knowing When to Teach: Learning from Teachable Prefixes in Multilingual Reasoning Distillation

Jaewon Cho, Minseo Yoon, Jinyoung Park, Jinwoo Seo, Jaewon Chu, Alice Oh — Fri, 06 Nov 2026 00:00:00 +0000

92% · New model 92%, Evaluation 35%

On-Policy Self-Distillation (OPSD) improves reasoning by training a student on its own reasoning, using the same model conditioned on reference solutions as a teacher. Applying OPSD to multilingual reasoning requires the teacher to provide useful feedback across languages, even when the student makes mistakes in intermediate steps. We observe a phenomenon we call Teacher Drift: as these mistakes accumulate, the teacher may follow the student's incorrect steps instead of correcting them. Our analysis shows that this shift tends to occur earlier in lower-resource languages. To account for these differences across languages, we propose Teachable On-Policy Self-Distillation (TOPSD), which determines where and whether teacher supervision should be applied to each reasoning trajectory. TOPSD uses teacher-student likelihood to estimate where drift begins and select a prefix for supervision. It then verifies whether the teacher can recover the correct answer from that prefix, applying distillation only to prefixes that pass verification. Experiments on multilingual mathematical reasoning across ten languages and three model scales show that TOPSD consistently outperforms conventional OPSD, particularly in low- and medium-resource languages.

AutoCompress: Automated Hardware-Guided Model Compression

Fri, 06 Nov 2026 00:00:00 +0000

92% · Efficiency 92%, Evaluation 68%, New model 45%

LLM pruning must preserve quality while meeting hardware- and workload-specific constraints on latency, throughput, and memory. Existing pruning workflows often target parameter counts, requiring practitioners to manually select model sizes that meet latency, throughput, and GPU memory limits. However, parameter count alone does not predict these metrics, which depend on the model architecture, workload, hardware, and serving engine. We introduce AutoCompress, a hardware-guided compression framework that directly incorporates such constraints into differentiable architecture search. AutoCompress jointly optimizes global and per-layer structure to produce heterogeneous architectures, guided by serving-engine cost models validated against end-to-end latency, throughput, and memory measurements. We evaluate AutoCompress on long-context local coding, batched datacenter serving, and latency-sensitive in-game interaction. Compared to the state-of-the-art Minitron pruning framework, AutoCompress improves average accuracy across eight benchmarks by 6.57, 2.35, and 0.57 percentage points on RTX 5070, RTX 5080, and H100, respectively. AutoCompress also compresses Nemotron 3.5 Super from 120B to 74B parameters for single-B200 BF16 serving at 128K context, retaining 95% of its mean benchmark score.

SwarmEvolve: Latent Agent Swarms for Evolutionary Discovery and Optimization

Qile Jiang, George Em Karniadakis — Fri, 06 Nov 2026 00:00:00 +0000

92% · Efficiency 92%, Evaluation 40%

LLM-guided evolutionary search uses AI agents to improve programs for mathematical and computational problems by iteratively refining solutions and combining high-scoring methods. However, the process is often slow because the LLM generates a long textual chain of thought for each round and must repeat this generation many times to maintain population diversity. We present SwarmEvolve, a framework in which hundreds of low-cost agents per round work in swarms, reason in latent space without decoding to text, and share information by passing KV caches. Each swarm studies the problem from a different angle. A novelty filter keeps the most distinct agent ideas by directly measuring the distances between the agents' hidden state vectors. Crossover of ideas happens naturally in latent space, as a summarizer agent receives the combined KV caches of all swarm members. On 21 problems in mathematical discovery and scientific computing, SwarmEvolve gets the highest average score and needs a median of only 3 to 8 minutes to reach the best scores that other state-of-the-art frameworks reach in 60 minutes. On several problems, SwarmEvolve with Qwen3 8B improves on the solutions that AlphaEvolve first found with frontier LLMs. Our work shows that agent swarms that communicate in latent space offer a fast, low-cost approach to program search for scientific problems.

On the Opportunities and Challenges of Prefill-to-Autoregressive Cross-Model Transfer

Zhenyi Shen, Junru Lu, Wei Liu, Jiangnan Ye, Jiazheng Li, Di Yin — Fri, 06 Nov 2026 00:00:00 +0000

91% · Evaluation 91%, New model 71%, Efficiency 62%

Prefill-to-Autoregressive Cross-model Transfer (PACT), a form of latent communication between language models, combines a larger model's prompt representations with a smaller model's efficient generation, offering a promising way to balance capability and inference cost. However, existing systems provide limited support for independently evaluating large-small model collaboration, and improved task performance alone does not establish whether the receiver benefits from the sender's information. We introduce PLatour, an open-source PACT framework that connects frozen models through lightweight, zero-initialized linear bridges trained directly with task rewards and supports hybrid-attention architectures. PLatour significantly outperforms open-source baselines on generation tasks and surpasses Mostik's reported results on shared mathematics and coding benchmarks using a substantially smaller sender. However, controlled experiments show that the gains on some tasks can be reproduced by uninformative sender signals, self-communication, or receiver-only training. We find that trainable bridges can act as parameter-efficient adaptation mechanisms without exploiting useful sender information. We therefore propose PACT-Check to quantify how much of a trained PACT system's gain genuinely comes from communication. Finally, we discuss the opportunities and challenges of PACT systems. Code is available at https://github.com/zhenyi4/platour.

BlockLock: Reproducible LLM Inference with Fixed Batch Geometry

Xiuyuan Qi, Pedro Reviriego, Javier Conde, Shanshan Liu — Fri, 06 Nov 2026 00:00:00 +0000

91% · Efficiency 91%, Evaluation 41%

Reproducible LLM inference requires that a request produces the same output regardless of the other requests in the batch. Existing solutions pay for this in different ways: batch-invariant kernels constrain every operator, and isolating prefill with fixed-shape decode ({our CoRun-style baseline}) gives up shared prefill compute and pads decode. We present BlockLock, which keeps shared batched prefill and pins only the numerical path of the protected request: it fixes the geometry that the request sees, with {two modes: BlockLock-Rows uses fixed-width rows and retains batched decode of all requests; BlockLock-Pack uses terminal-anchored packing with an isolated dummy segment, then extracts the protected request's KV state for target-only decode. Under a stated per-deployment contract, both modes preserve the target's output bitwise when its companions change.} Across five {dense} models (Qwen3-8B/14B, Llama-3.1-8B-Instruct, and Gemma-3-4B/12B-IT) and three single-GPU deployments (NVIDIA A6000, A100, and H100), we evaluate our approach on both tinyBenchmarks and BFCL{; we also include initial tensor-parallel experiments on two NVIDIA A6000 GPUs. Both modes preserve target sequences and Top20-logit bits in the reported tests.} At equal delivered work, {BlockLock-Rows} reduces latency by 53.5--55.0\% and improves throughput by 115.2--122.2\% relative to vLLM Batch-Invariant; compared with batch-invariant kernels, it needs no custom kernel implementations and constrains only the protected requests.

When Verification Is Not Agreement: Cross-Agent LLM-to-Dafny Formalization

Priyanath Maji — Fri, 06 Nov 2026 00:00:00 +0000

91% · Evaluation 91%

A verified Dafny postcondition certifies that a formula holds, but it does not guarantee that the result was reached through genuine independent reasoning. We introduce a cross-agent formalization protocol where a Specifier model translates clinical trial criteria into a Dafny predicate, and an Implementer model writes an operational method checked against it. A same-model control establishes that even with no capability gap, independent re-extraction from identical text agrees only 59.3% of the time. Evaluating four model configurations on 96 clinical trial documents, we find that agreement is highly sensitive to how model capability is allocated across the two roles: the GPT-4.1-nano/GPT-4.1-mini configuration achieves 53.0% agreement when the stronger model is the Implementer, but only 14.3% when the roles are reversed. The strongest tested configuration, GPT-5.6-luna/GPT-5.4-mini, improves agreement to 34.4% but remains below the same-model baseline and is dominated by direct contradiction (42.7%). We also detect a critical verification artifact in which a weaker Implementer satisfies the postcondition by invoking the Specifier's predicate directly; after structural auditing, corrected agreement falls from 14.3% to 13.2%. Our results show that verification success alone is insufficient evidence of independent cross-agent agreement, and that model capability does not guarantee agreement when formalizing the same source independently.

How Do AI Agents Waste Your Money?

Huimin Zeng, Yuxuan Lu, Jesse Cai, Eric Hanchen Jiang, Di Wu, Po-Nien Kung — Fri, 06 Nov 2026 00:00:00 +0000

91% · Evaluation 91%, Efficiency 72%, Agents 62%

AI agents may consume massive tokens across a wide range of tasks, but how many of those tokens are really necessary? Although prior work has shown that agents can incur unexpectedly high token costs, we still lack a systematic way to identify and measure token waste. In this work, we introduce a unified framework for measuring token waste across heterogeneous agent systems. Our framework records and normalizes complete agent trajectories, identifies removable spans using an LLM-as-a-Judge (LLMaJ), and maps the identified text spans back to the original trajectory for token-level measurement. We validate the identified waste through direct interventions by removing marked spans, resuming agent execution, and measuring changes in task outcomes and subsequent actions. In online pruning experiments, we remove 41-46% of the model's reasoning while largely preserving task performance across two distinct agent benchmarks. Using this framework, we analyze agents across multiple settings. Our analysis shows that token waste has a system-level structure that aggregate token usage alone cannot reveal. First, waste varies more across tasks and agent settings than across models. Second, reasoning and visible execution can exhibit different waste patterns. Third, different harnesses can introduce distinct sources of inefficiency even under the same model and task. Overall, our framework transforms token waste into an actionable optimization signal, enabling researchers and practitioners to train more efficient models, design better reasoning strategies, and optimize agent harnesses toward reducing unnecessary computation rather than merely reducing token usage.

When Scores Conflict with Preferences: Pre-Alignment Diagnosis and Adaptive Correction of Constraint Drift in DPO

Ping Liu, Yan Yan — Fri, 06 Nov 2026 00:00:00 +0000

91% · New model 91%, Evaluation 48%

Direct Preference Optimization (DPO) is increasingly used in settings where pairwise preferences coexist with pointwise criteria such as response length, answer correctness, or safety scores. When the two disagree, standard DPO treats pairs whose preferred output scores worse on the pointwise criterion (conflict pairs) in the same way as the remaining (matching) pairs. We characterize a systematic failure mode, constraint drift: for sufficiently small updates, conflict pairs tend to increase constraint cost while matching pairs tend to repair it, and the net change depends on their balance. This balance yields a closed-form pre-alignment diagnostic, $\rho^{*}_{\mathrm{eq}}$, computable from dataset statistics, that indicates whether a dataset is self-correcting or requires intervention. Building on this diagnostic, we propose Conflict-aware DPO (CoDPO), which down-weights conflict pairs with a strength set from $\rho^{*}_{\mathrm{eq}}$ rather than tuned, and adds an adaptive margin shift that addresses residual drift during training. Empirically, drift grows monotonically with conflict exposure on the conflict-dominated domains, and the intervention decision given by $\rho^{*}_{\mathrm{eq}}$ matches the observed drift on every domain, including one that it identifies as self-correcting. On all four domains of the main comparison, CoDPO reduces drift while largely preserving the preference learning of vanilla DPO. Compared with hard filtering and recent constrained and multi-objective baselines, CoDPO also achieves a more favorable drift-preference trade-off.

DelegateEval: Evaluating Platforms for Personal External Service Agents

Jichao Li, Jinrong Lu, Manjiang Yu, Zhixuan Liu — Fri, 06 Nov 2026 00:00:00 +0000

91% · Evaluation 91%, Agents 72%

AI agents that work for an individual over time may also help that individual serve external clients. We define a personal external service agent as an agent authorized and continuously managed by a specific individual to accept client requests, deliver work, and provide ongoing support using that person's knowledge, experience, professional methods, and service boundaries. Such agents could allow individuals to serve more clients with less direct involvement and generate economic value at scale. We propose DelegateEval, a platform-level evaluation framework for agent systems that may combine large language models, reusable skills, protocols such as MCP, and executable tool environments. It covers eight dimensions: ease of creation, service discovery and matching, delivery quality, expertise inheritance, service continuity, learning and improvement, reliability and governance, and efficiency and value. These dimensions assess whether a platform enables people with different backgrounds to turn their expertise into ongoing services that deliver verifiable work, develop capabilities through experience, and remain economically feasible. We present the framework and an experimental design. Complete datasets, executable environments, a unified scoring protocol, and cross-model results remain to be developed and empirically validated.

Follow the Policy, Fail the Benchmark: Auditing and Revising Policy-Based Decision Benchmarks

Yanru Xiao, Yifei Cai — Fri, 06 Nov 2026 00:00:00 +0000

90% · Evaluation 90%

Policy-based decision benchmarks evaluate whether systems apply written rules to financial and operational cases. Yet a rule-following response can be marked wrong when the benchmark misstates the required decision. We audit nine public policy-based decision benchmarks through source tracing, program execution, and controlled checks. We find unsupported targets, reference errors that survive replay or propagate as solved steps, and evaluators that reject acceptable answers or credit incomplete decisions. To measure the scoring impact, we rescore unchanged model responses. On the 96 LegalBench SARA numeric cases, correcting the scorer raises acceptance by more than 80 percentage points for each of three models, without changing references or responses. In RuleArena Tax, restoring a required zero floor changes 27 of 300 references and narrows an observed model gap. A controlled federal-exemption task with explicit ownership and eligibility assumptions also shows that errors in supplied intermediate answers can affect final decisions. We present BAR (Benchmark Audit and Revision), a versioned collection of audit evidence and revisions for reevaluation.

Latency-Aware Adaptive Best-of-$N$

Alon Jacobson, Qian Xie, Peter I. Frazier — Fri, 06 Nov 2026 00:00:00 +0000

90% · Efficiency 90%, Evaluation 65%

Test-time compute (TTC) improves language model responses by spending additional computation at inference time, and the right amount of compute depends on the prompt. Adaptive TTC methods allocate compute based on signals of response quality such as observed rewards, often by stopping once the signal exceeds a fixed threshold, and generally measure cost by the number of generations or tokens. But in online serving, latency is also a cost: it grows with server load, so a method using a fixed threshold spends too little compute when generation is fast and too much when it is slow. For Best-of-$N$ sampling, a widely used TTC method, we propose a two-stage adaptive algorithm that draws an initial batch, then uses the observed rewards and the batch's latency to decide whether to draw a second batch and at what size by comparing expected improvement against predicted cost: it draws one when the first batch's best reward is below a threshold that decreases with the first batch's latency, without explicitly measuring server load. We map each reward to its value under the cumulative distribution function (CDF) of the prompt's reward distribution, so the same tradeoff parameters can be reused across tasks; a regression fitted offline on other prompts predicts a distribution for the CDF value of the first batch's best reward, giving the expected improvement in closed form. We evaluate the method by simulation on a preference-based alignment benchmark, measuring generations-equivalent cost: generations plus latency, converted to generations. The method reaches Best-of-$N$'s reward at lower generations-equivalent cost, and conditioning on the measured first-batch latency saves at least as much of that cost as fixed-threshold (latency-unaware) baselines, and more when latency is highly variable, when it grows with batch size, and when the second batch's latency is highly predictable from the first's.

AttnSR: Attention Step Reward for Long-Horizon Agentic RL

Mengsong Wu, Tianchen Zhou, Zhaochen Su, Zhenhua Liu, Zhe Cui, Aimin Zhou — Fri, 06 Nov 2026 00:00:00 +0000

90% · New model 90%, Agents 68%, Efficiency 53%

Long-horizon LLM agents make multi-step decisions, yet outcome-reward-only reinforcement learning provides little direct supervision for individual steps. We introduce AttnSR, a step-level reward augmentation method that derives reward allocations from policy attention extracted during training replay, without a separately trained reward model or additional rollouts. AttnSR traces backward through the policy model attention graph to estimate the relevance of earlier steps to the final response. The terminal step, measured relative to a running baseline, determines whether these rewards reinforce or penalize the selected steps. In a controlled Qwen2.5-7B-Instruct study, AttnSR consistently outperforms the PPO baseline across all evaluated checkpoints on ALFWorld, raising final held-out success from 91.2% to 97.1%. The ablation experiment shows that the benefit depends on how intermediate rewards are assigned, rather than simple accumulation of step rewards over time. AttnSR also improves ScienceWorld at every checkpoint while cutting environment interactions by 29.9%. These results support policy attention as a practical prior for step-level reward allocation in long-horizon agentic RL. Code is available at https://github.com/fairyshine/AttnSR .

CardioLM: A multilingual suite of small language models for the cardiology domain

Fri, 06 Nov 2026 00:00:00 +0000

90% · New model 90%, Evaluation 87%

Background: Cardiovascular care generates large volumes of clinical free-text documentation in local languages. However, clinical natural language processing (NLP) resources remain concentrated in English, and no existing benchmark is cardiology-specific, multilingual, and broad in entity coverage. Methods: We present two contributions. The first is CardioCCC, a corpus of 508 cardiology case reports originally published in Spanish and extended to Czech, Dutch, English, Italian, Romanian, and Swedish through machine translation, annotation projection, and clinician review, with annotations covering diseases, symptoms, procedures, and medications. The second is CardioLM, comprising CardioBERTa, a set of language-specific encoders obtained by continued pretraining on cardiology and biomedical corpora, and CardioNER, a set of named entity recognition (NER) models fine-tuned on CardioCCC. Results: Cardiology-specific pretraining reduced pseudo-perplexity by up to a factor of eight, yet changed strict span-level F1 by at most 0.04, with benefits concentrated in languages whose backbones lacked biomedical coverage. The jointly trained multilingual XLM-RoBERTa-large models achieved the best strict F1 in nearly every language and entity type, exceeding the strongest monolingual system by up to 0.13 F1. A large share of residual errors concerned boundary detection: relaxed matching recovered 0.10-0.23 F1 for disease, procedure, and symptom, but only 0.01-0.07 for medication. Clinicians rated CardioNER well above a rule- and dictionary-based baseline, with the mean rating across entity types rising from 26% to 73%. Conclusions: Specialty-specific pretraining is not the decisive factor for downstream entity recognition, whereas multilingual joint training yields the strongest results in nearly all settings and a simpler deployment path for multi-country clinical information extraction.

Towards interpreting any residual stream direction by scaling activation inversion

Fri, 06 Nov 2026 00:00:00 +0000

90% · New model 90%, Evaluation 64%

To interpret a direction in a language model's residual stream, we can look for text that strongly activates it. But when such text is found by scanning a corpus, it is limited to what that corpus contains. We develop Maximally-Activating Example Metamodels (MAEMs), language models which map residual stream directions from a subject model to text exemplars that strongly activate those directions. Our objective is to maximise the cosine similarity of a supplied residual stream direction with the activations of the text generated by the MAEM. This label-free RL objective helps to scale MAEMs on diverse sources of residual stream directions. Empirically, we find the resultant metamodel produces informative exemplars: a Qwen3.6-27B MAEM inverts held-out activations better than corpus search or a natural-language autoencoder (NLA), beats corpus search (at 10M tokens) on a majority of held-out languages/domains, and activates 89% of held-out SAE features. It also identifies concepts behind steering vectors about as well as NLAs, recovers the payloads of LoRA backdoors more often than corpus search, and can even verbalise some model workspace inner monologue. Unlike explanations produced by other metamodels (e.g. NLAs), MAEM exemplars can be verified with a forward pass on the subject model, making the MAEM objective a grounded alternative for scaling the training of metamodels.

Turns Matter: Advancing Streaming Video RL via Turn-Level Credit

Shuai Dong, Kanghui Tian, Linghan Chen, Wenqi Shao, Hongrui Wu, Haowen Hou — Fri, 06 Nov 2026 00:00:00 +0000

90% · New model 90%

Streaming video understanding requires models to repeatedly decide whether to respond or continue observing. Recently, Group Relative Policy Optimization (GRPO) has been applied to this setting, with existing approaches emphasizing reward design while retaining a single trajectory-level advantage across turns. Such shared credit can reinforce redundant replies after earlier successes and discourage useful follow-ups after earlier failures. We propose TurnWise, a turn-level GRPO framework that aligns learning with individual streaming decisions. TurnWise addresses the mismatch between trajectory-level feedback and turn-level interaction, promoting informative responses and reducing redundancy. Specifically, turn-level credit assignment uses silence-completed prefix scoring to construct group-relative advantages from each decision's remaining return, separating its learning signal from earlier rewards. Turn-mean policy optimization averages token losses within each turn and then across turns, giving response and waiting decisions equal weight regardless of output length. These complementary designs improve streaming policy learning without a learned critic or additional rollouts. Extensive experiments show our method outperforms prior RL approaches for streaming video understanding.

★ SRRM: A Retrieval-Reconstruction Benchmark for Evaluating Long-Context Memory in Small Language Models

Christopher Alberto Hamonangan — Thu, 05 Nov 2026 00:00:00 +0000

99% · Evaluation 99%

Small Language Models (SLMs) are increasingly deployed on consumer-grade hardware, yet their ability to preserve information across progressively longer contexts remains insufficiently understood. Existing benchmarks primarily evaluate retrieval-oriented tasks, providing limited insight into whether successful retrieval reflects comprehensive long-context memory preservation. To address this limitation, this paper introduces Simple Retrieval-Reconstruction Memory (SRRM), a lightweight and fully reproducible benchmark that jointly evaluates Targeted Retrieval and Global Information Reconstruction. This study also proposes the Memory Preservation Ratio (MPR), a benchmark-level metric that quantifies the relationship between retrieval and reconstruction performance. SRRM was evaluated using a synthetic dataset on three Qwen2.5 models (0.5B, 1.5B, and 3B parameters) across context loads ranging from 10 to 100 factual statements. Experimental results show that reconstruction performance deteriorated substantially faster than retrieval accuracy as contextual information increased, indicating that retrieval alone is insufficient to characterize long-context memory preservation. Among the evaluated models, Qwen2.5-1.5B achieved the highest retrieval accuracy (86.7%), reconstruction performance (45.7%), and MPR (0.53), outperforming the larger Qwen2.5-3B model (72.2%, 18.5%, and 0.26, respectively). These findings demonstrate that increasing parameter count does not necessarily improve long-context memory preservation. Overall, SRRM provides a reproducible and extensible benchmark for systematically evaluating long-context memory while revealing complementary memory behaviors that remain hidden under retrieval-only evaluation frameworks.

★ SlackBench: Benchmarking Agents on Collaborative Projects Grounded in Real Code Repositories

Thu, 05 Nov 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 91%, Hallucination 53%

AI Agents are becoming digital coworkers that can browse the web, write code, run experiments, and assist humans over extended horizons. Yet existing benchmarks fall short of real knowledge work: many rely on synthetic data or isolated tasks, missing the complex, multi-party dynamics of collaborative projects, where critical information is scattered across conversations, code, and evolving decisions. Even simple workplace questions may require locating relevant channels and direct messages, recovering context from long discussions, inspecting repository state, reconciling stale or conflicting evidence, and synthesizing across sources. We introduce SlackBench, a benchmark for evaluating agents on realistic research and engineering workflows that require joint reasoning over workplace communication and evolving repository state. Grounded in privacy-preserving reenactments of real academic research projects, SlackBench provides reproducible environments with realistic multi-party dialogue, evolving decisions, and corresponding code changes grounded in GitHub repositories and Google Docs. It contains 140 queries spanning heterogeneous-source agentic search, abstention on unanswerable or false-premise questions, goal-directed summarization, and circumstance inference over long conversations. These tasks are not reducible to static corpus question answering: agents must determine which conversations matter, traverse threads, inspect repository state, and synthesize evidence across communication and code artifacts. Across 33 frontier models and agent harnesses, the best system scores around 80% and most agents score below 66%, highlighting substantial headroom for improvement. SlackBench thus provides an extensible framework for evaluating agents in collaborative work environments. The SlackBench dataset is available at https://huggingface.co/datasets/HuanzhiMao/SlackBench .

★ Phonological Knowledge of Chinese Characters in Language Models

Xiaozhe Ji, Renfen Hu, Hai Hu — Thu, 05 Nov 2026 00:00:00 +0000

99% · Evaluation 99%

How much phonological knowledge do language models (LMs) possess about Chinese characters, and how is this knowledge represented and used? We introduce PinyinBench, a controlled benchmark covering homophones, phonological features (initials, finals, and tones), and polyphonic characters. We investigate 17 text-only and multimodal LMs through three complementary approaches: behavioral evaluation, layer-wise probing, and activation steering. Behaviorally, models demonstrate substantial knowledge of initials and finals, with a frontier model (DS-v4-flash) scoring 90%+ in homophone judgment, when two characters share no phonological features. However, its accuracy drops sharply to 72% when two characters differ only in lexical tone. Probing reveals that tone is highly decodable from embeddings and improves modestly across layers, whereas segmental information starts less decodable and improves substantially with depth. Activation steering shows that segmental representations can systematically influence generation, while tone exhibits much weaker effects. Furthermore, audio-bearing training improves phonological decodability, including tone, without corresponding improvements in steering effectiveness. Together, these findings reveal distinct representational and functional profiles for segmental and tonal information, with tonal information much harder to access and manipulate, demonstrating the necessity for combined evaluation, probing, and causal intervention for understanding linguistic knowledge in LMs.

★ CARTBench: Rubric-Based Evaluation of Shopping Advice for Complex and Ambiguous Requests with Timelines

Yuqin Dai, Zhanwei Zhang, Honghui Yang, Zhouheng Yao, Yueqi Zhang, Jingyu He — Thu, 05 Nov 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 42%, Hallucination 36%

Large language model (LLM) shopping assistants now recommend products in multi-turn conversations. However, reliable evaluation faces Data scarcity and Difficulty Standardizing Criteria. To address these obstacles, we introduce CARTBench for CART (Complex and Ambiguous Requests with Timelines). For Data scarcity, we curate cases from real commercial shopping needs across 17 retail domains, pairing timestamped dialogue with 1,603 product records. For Difficulty Standardizing Criteria, we develop case-specific rubrics guided by Ideal Response Principles, distilled from excellent replies manually selected from successfully completed orders. These principles also guide Principle-Guided CART Prompting, our structured recommendation strategy. The experimental benchmark contains 1,328 expert-annotated criteria across Core, Excellence, and Safety. To answer the CART question, we evaluate 7 flagship commercial API models, including GPT-6.1-Sol and Claude-Opus-5, and 35B+ parameters open-weight models. In the first full-suite evaluation under the frozen rubrics, 3 of 10 models exceed 90% on Core, while none exceeds 45% on Excellence. Therefore, we identify an understanding-to-execution gap between required behavior and high-quality purchase advice. To examine these limitations, we diagnose shared failures in product evidence and unresolved needs, linked to changing requirements and decisive unknowns. Code and data are available at https://github.com/Da1yuqin/CARTBench.

★ NaijaFacts: A Framework for Disentangling Language from Knowledge in Facts about Nigeria

Zainab Afolabi, Nikita Kozodoi, Jack Butler, Babs Khalidson — Thu, 05 Nov 2026 00:00:00 +0000

99% · Evaluation 99%

Multilingual benchmarks for African languages mostly translate generic questions into local languages, measuring linguistic competence but not whether models know facts about Africa. We introduce \naija, a benchmark of factual and statistical questions about Nigeria, authored by native speakers and grounded in authoritative official statistics, with every question available in English, Nigerian Pidgin, \yoruba, Igbo, and Hausa. Answering similar questions for the UK and US yields a country by language matrix that disentangles a language effect (the model knows the fact but degrades in a low-resource language) from a knowledge effect (the fact is absent in every language). Across eight frontier and Nigeria-adapted models, the knowledge effect dominates: English-query accuracy is 49--50\% for the US/UK but 20\% for Nigeria. The language effect is smaller and tracks web representation, degrading from English to \yoruba{} (9 points on Nigeria, up to 24 on US/UK questions). We find scale does not help: the strongest models reach 66--79\% on the UK and the US but none exceed 31\% on Nigerian statistics. Extending to all 54 African countries with our best performing frontier model shows the same profile continent-wide, 96\% on culture and 98\% on politics against 26\% on statistics, so the gap is not specific to Nigeria.

★ LLMs Don't Know What They Know: Benchmarking Temporal Self-Knowledge for Financial Backtesting

Lingfei Qian, Yan Wang, Xueqing Peng, Yupeng Cao, Yuechen Jiang, Zhuohan Xie — Thu, 05 Nov 2026 00:00:00 +0000

99% · Evaluation 99%, Hallucination 56%

Financial backtests of large language models (LLMs) are meaningful only if the model did not already know the outcomes it is asked to predict. Practitioners guard against this by trusting a vendor's stated knowledge cutoff, by asking the model what it knows, or by instructing it to answer as of a past date. To verify whether any of these safeguards can be trusted, we introduce FinDendro, a benchmark of 2,226 dated U.S. facts from 1999 to 2026, covering monthly macroeconomic series, month-end stock prices and Federal Reserve decisions. Each fact yields four prompts: a value question, Recall and Predict questions that pair each candidate value with a claim of knowing or guessing without a declared date, and a Predict question with a historical declared date. Controls for persistence and dominant outcomes keep chance accuracy near 0.27. Results show that for most frontier models, self-reports understate knowledge. GPT-5 answers 76% of controlled facts correctly but claims to know 38%, and answers 61% of the facts it says it does not know. Knowledge claims can persist beyond the measured knowledge boundary, by up to seven quarters, and a declared date removes most of the claims of prior knowledge while leaving accuracy unchanged. Backtests inside a model's training window therefore cannot be certified leak-free by self-report, instruction or stated cutoff. These findings motivate behavioral measurement of temporal knowledge boundaries, and FinDendro provides a benchmark and procedure for selecting backtest periods based on demonstrated factual knowledge.

★ Before You Continue: Evaluating Time-Sensitive Safety Awareness in Conversational Assistants

Peidi Dong, Yunze Xiao, Yu-Lin Tsai, Yichen Zheng, Heyuan Huang, Weihao Xuan — Thu, 05 Nov 2026 00:00:00 +0000

99% · Evaluation 99%

Large language model (LLM) assistants increasingly help users with everyday activities that involve time-dependent constraints. However, users may not mention these constraints in later requests, requiring assistants to recognize and adjust their advice accordingly. To evaluate such capability, we introduce TASC (Temporal Action Safety in Conversation), a benchmark that evaluates whether LLMs apply time-dependent constraints when responding to conversational requests. TASC consists of 3,418 inputs derived from 128 templates, along with a separate set of 33 scenarios designed to compare models' ability to calculate elapsed time with their ability to apply that information. Across 13 models, assistants account for time-dependent constraints in only 30.5% of cases where the proposed action violates a rule, while rarely warning users unnecessarily. More importantly, models fail to apply these constraints in 62.1% of cases despite correctly calculating the elapsed time. Explicitly providing the timing resolves only about half of these failures, revealing a gap between temporal reasoning and its application in conversational advice.

★ Towards Robust Long-Lived Agents: Benchmarking Experience Lifecycles in Evolving Environments

Shijue Huang, Hangyu Guo, Hengheng Zhang, Guo An, Zhitao He, Qi Tian — Thu, 05 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 78%

As agents become more capable, they are moving beyond one-off tasks toward sustained assistance across successive user goals. Such assistance requires agents to discover how unfamiliar environments work, reuse what they learn across tasks, and keep that experience current as environments evolve. Existing benchmarks increasingly study experience reuse and adaptation, but provide limited insight into which parts of accumulated experience agents revise, retain, or lose. To address this gap, we introduce ReScopeBench, a benchmark for the experience lifecycle across 18 interactive environments. It distinguishes structural experience about how an environment is organized and accessed from rule experience about reusable principles and constraints that guide decisions across tasks. To study how experience is acquired, reused, and revised, we propose an experience-first construction approach. It builds task streams around shared experience and introduces silent environmental changes. We evaluate task performance and track changes in saved experience throughout each stream. Across 20 models, we find that retained experience improves task performance, but agents struggle to keep it current as environments change. Even when the required experience is correctly recorded, agents do not always complete tasks successfully. These findings identify selective maintenance and effective use as complementary challenges for long-lived agents.

★ ViBench 2.0: A Benchmark for Long-Horizon Software Building

Thu, 05 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 86%, Efficiency 40%

On Replit, our agentic engineering platform, users now build entire products without writing code, through successive feature requests to an agent. Yet existing coding benchmarks evaluate single, fully specified tasks, and many are near saturation. We introduce ViBench 2.0, a benchmark for long-horizon software building that we use to select our platform's models. Its 17 full-stack web products, modeled on widely used software, grow over 241 feature requests and are judged only when finished, as a user would judge them: a browser agent runs 544 test plans, and a build passes only if every plan does, under a grading protocol validated to keep this strict score reliable. Evaluating nine models, we find a reliability gap: the three strongest pass up to 98% of test plans yet deliver a fully working product in at most 65% of builds, and no open-weight model delivers one. The gap is closing fast: across three releases of one model family, pass@1 on the public products rises from 3% to 53% while the cost per build falls by more than half. We release eight products with the harness and grader, hold out nine, and report lessons from production use.

★ PersonaCodeBench: Evaluating Coding Agents in Persona-driven Human-Interactive Sessions

Haixin Wang, Jianheng Hou, Hongyu Shen, Wenzhe Fan, Wei-Chieh Huang, Qianyu Zhu — Thu, 05 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 37%

Coding agents must deliver correct software while respecting users' expertise, preferences, and working conditions. Existing benchmarks increasingly incorporate interaction, yet personalized requirement fulfillment remains underexplored. We introduce PersonaCodeBench, which turns user characteristics into verifiable obligations in persona-driven multi-turn coding sessions. Our persona pool spans roles, technical abilities, interaction preferences, and practical constraints, yielding task-grounded evolved requirements for implementation, permissions, testing, explanation, and delivery. At runtime, a controller governs user simulator responses and requirement disclosure, using online requirement evaluation to provide feedback and advance the interaction. Requirements emerge progressively, with their content, disclosure conditions, and acceptance criteria fixed before evaluation. A completion verifier and a personalization verifier separately assess native correctness and persona fulfillment. Joint pass requires both adherence to the declared budget and compliance with validity conditions. We evaluate all major frontier agents, establish a comprehensive leaderboard, and conduct detailed experimental analyses to validate effectiveness. PersonaCodeBench supports reproducible evaluation of coding agents' ability to meet explicit, personalized user expectations alongside coding objectives.

★ TandemBench: Benchmarking Decision Models for Multi-Agent Collaboration

Baohang Li, Xiaocheng Feng, Bing Qin — Thu, 05 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

Decision models choose among predefined options in tasks such as routing and action selection. In multi-agent collaboration, a decision determines how members divide work, exchange resources, and act in parallel. We introduce TandemBench, a benchmark of 1,200 questions in multi-agent cooking scenarios where coordination choices affect goal completion. Models receive complete global information and candidate team plans, with reference answers established through execution and exact search. Models select an optimal team plan and, in paired variants that retain or remove it, judge whether better arrangements exist beyond the supplied options. We evaluate 40 native decision models and compare them with 10 general-purpose language models. The best native decision model achieves 62.50% decision accuracy. Native decision models remain close to chance in optimality verification. Paired tests reveal that some report better alternatives regardless of whether an optimal plan is already present. These findings show that team-plan decision accuracy alone can conceal systematic errors in judging whether the supplied options contain an optimal arrangement. The dataset and evaluation code are publicly available at https://github.com/lililibaohang/TandemBench.

★ ProcessPrefBench: Evaluating Cross-Task Transfer of Individual Process Preferences from Naturalistic Human-AI Interaction

Yiwen Yang, Guanming Liu — Thu, 05 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 89%

In natural human-AI dialogue, users' corrections, requests, and self-descriptions carry reusable signals about how they want work to be done. These signals reveal individual process preferences, user-specific standards for evidence, depth, analytical focus, procedure, and delivery. We ask whether a preference revealed in one dialogue can shape an assistant's later answer when the task does not restate it. ProcessPrefBench uses three months of natural dialogues from two knowledge workers to build 66 source-target cases separately judged for source evidence and target applicability. Sources are preselected, so the benchmark tests application rather than retrieval. We test three models with target-only, full-history, and distilled-preference prompts. Target-specific 0-2 rubrics score preference realization, and a task gate flags severe failures. Compared with target-only prompting, full-history and imperative distilled-preference prompting raise preference realization by 0.204 and 0.244 points on the 0-2 scale, respectively, with no reliable advantage from distillation. Fabricated-evidence failures rise from 5.8% under target-only prompting to 10.1% under full-history and 10.4% under imperative prompting. ProcessPrefBench highlights a tension in learning from dialogue, where drawing on prior interactions can improve preference realization yet also invite fabricated evidence. We hope it helps build assistants that honor a user's process preferences without sacrificing factual reliability.

★ Accelerating Long-Context Decoding in Large-Scale Hybrid LLMs Without Retraining

Jo\~ao Eduardo Batista, Emil Vatai, Mohamed Wahib — Thu, 05 Nov 2026 00:00:00 +0000

98% · Efficiency 98%, Evaluation 46%

The full-attention layers of large language models (LLMs) make every generated token more expensive than the previous one, slowing models down on long-context tasks. Hybrid models reduce this cost by replacing most full-attention layers with linear-attention layers, but the first still dominate decoding at long contexts. We accelerate these remaining layers without any training or weight changes using combined sparse attention: each layer attends to a sliding window of recent tokens, a pinned prefix, and a fixed budget of most relevant tokens in the context. The window size and budget are found on a per-layer basis using a multi-objective search algorithm. Across six hybrid models from three families, from 4B to 397B parameters, the selected configurations preserve their baseline model's accuracy on 21 long-context datasets. Decoding 64k/256k tokens becomes up to 1.26x/2.22x faster on Qwen3.5-4B, 1.20x/2.23x on Qwen3.5-122B-A10B, and 1.16x/2.06x on Ornith-1.0-397B.

★ A Systematic Benchmark for Evaluating Multimodal Personalized Reasoning of Vision-Language Models

Dohyun Kim, Junyong Lhim, Kwanyong Park — Thu, 05 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

People often make decisions by holistically reasoning over heterogeneous forms of personal information. In this paper, we study the relatively underexplored capability of multimodal personalized reasoning in multimodal large language models (MLLMs). To this end, we introduce \texttt{MPRBench}, a novel and comprehensive benchmark specifically designed to evaluate personalized reasoning across a wide range of tasks and real-world challenges. \texttt{MPRBench} consists of 12 sub-tasks and additionally supports interpretable analysis of representative error cases. Through extensive experiments, we find that current personalized MLLMs still struggle with personalized reasoning due to several key challenges, including inaccurate recognition of personalized concepts and sensitivity to irrelevant personal information. We hope that MPRBench will stimulate further research on personalized reasoning in MLLMs. Code and data are available at https://github.com/kkdh15/MPRBench

★ Twists and Turns: A Graph-Based Benchmark and Framework for Long-Term Memory in Exploration Tasks

Chongyuan Dai, Yaling Shen, Zirui Song, Junfeng Liao, Jinpeng Hu, Xiuying Chen — Thu, 05 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

Long-term problem-solving processes, such as scientific research and software development, proceed along an evolving trial-and-error trajectory toward a central goal. We term these \textit{exploration tasks}, which require models to remember not only facts but also how solutions are tested against intermediate outcomes and early attempts are continually revised or revived. However, existing long-term memory research focuses on factual recall and updates, overlooking the interconnected nature of exploration. We observe that exploration processes naturally form graphs, with branching approaches linked by revisions and reconsiderations. Motivated by this, we propose \textsc{LExMem}, a graph-guided benchmark designed to evaluate \underline{\textsc{Mem}}ory in \underline{\textsc{L}}ong-term \underline{\textsc{Ex}}ploration tasks. We model exploration processes as graphs and introduce an evolutionary mechanism to increase their structural complexity, yielding 224 dialogues and 2,018 question-answer instances derived from graph structures. Building on the same insight, we propose \textsc{LExGraph}, a graph-based memory framework that incrementally organizes dialogues into interconnected memory nodes and retrieves evidence through hybrid retrieval and neighbor expansion. Extensive experiments demonstrate that \textsc{LExMem} poses a significant challenge to existing retrieval- and memory-based systems, whereas \textsc{LExGraph} achieves superior performance and generalizes effectively to external benchmarks. Codes and datasets are available at https://github.com/mbzuai-nlp/LExMem.

★ AmbiHTMLBench: A Benchmark for HTML-based Visually Rich Document parsing

Boyu Niu, Yingjie Xiong, Shurui Li, Linhan Pang, Yuntong Zhang, Han Wang — Thu, 05 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

Traditional document parsing systems use Markdown as their output representation, which is effective for linear text but does not fully preserve fine-grained spatial structure, table geometry, visual hierarchy, or rendering semantics. We show that HTML is advantageous for heterogeneous and layout-sensitive documents whose representations must remain editable, inspectable, and reusable, and introduce AmbiHTMLBench to systematically study this setting. In this benchmark, given a real-world document, a model must recover executable HTML that preserves its content, organization, and visual appearance. The benchmark contains 1,115 diverse pages with structured annotations and evaluates predictions in two different ways: The objective protocol aligns ground-truth and predicted rendered blocks and measures table fidelity, page-level text error, and a two-dimensional layout consistency; a complementary reference-anchored protocol measures perceived fidelity by aligning LLM judges with human preferences. Across ten primary multimodal systems, the best objective score is 0.8092, while the subjective winner is different, demonstrating that automatic decomposition and human-perceived fidelity capture complementary properties. We further fine-tune a small-parameter expert model and find that task-specific adaptation yields good performance on document-to-HTML reconstruction task. These results position AmbiHTMLBench as a diagnostic testbed for building smarter and more logical document parsing systems.

★ LibriSpeech-Real-QA: A Benchmark for Streaming Question Answering

Zewen Zhan, Hui Lu, Xueyuan Chen, Huimeng Wang, Yikai Huang, Xixin Wu — Thu, 05 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

Most speech large language models (SLMs) still process speech in an offline manner. They only start responding after they receive the complete speech segment. However, many real-world applications require streaming speech interaction, for which the audio continuously flows in and the machine is required to answer questions based on the so far received audio chunks. For example, a user may ask the system to continuously count how many times a person's name has been mentioned while they are speaking. We observe a lack of benchmarks and evaluation protocols for real-time speech interaction. Motivated by this, we propose LibriSpeech-Real-QA, a dataset and a benchmark for real-time speech interaction. We construct different types of real-time question-answer pairs from LibriSpeech to evaluate whether speech LLMs can identify target words in streaming audio. We evaluate several popular real-time commercial models and find that they still tend to respond only after the speech turn has finished. We demonstrate that we can enable a pretrained SLM to perform this task, where we explore several modeling designs and establish a solid baseline for future research.

★ Tools Need Not Speak in Text: Latent Context Compression for Agents

Yuetong Liu, Zhehao Zhou, Yitong Zhang, Haorui Chen, Jia Li — Thu, 05 Nov 2026 00:00:00 +0000

98% · Agents 98%, Efficiency 96%, New model 72%

Large language model (LLM) agents typically append raw tool outputs to their context. Over multi-step interactions, long and noisy observations incur substantial context overhead and dilute the agent's attention to task-relevant information. Existing token pruning and summarization methods remain constrained by discrete text representations, limiting the achievable compression ratios. In this work, we challenge the prevailing assumption that agents must consume tool observations as explicit text tokens. To this end, we propose Coca-Cola, a continuous latent context compression model for LLM agents. Coca-Cola compresses tool observations into latent tokens and embeds them into the input embedding space of the agent. We further introduce Action-Preserving Training, which aligns the agent's next-action distributions under full and compressed observations. This trains the compression model to learn action-sufficient latent tokens that retain decision-critical information. Experiments with web-search agents on four QA benchmarks and two model families show that Coca-Cola reduces the agent's token consumption by 70.01%-89.24% and end-to-end runtime by 24.45%-55.97%, while consistently improving task performance. These results suggest that Coca-Cola offers a promising approach for context compression in agents. Our code is available at https://github.com/THU-Agent/Coca-cola.

★ SecureGARD: Dynamic Ontology-Guided Governance for Secure Tool-Using LLM Agents

Guoqin Xiong, Shengtao Wen, Xiang Chen, Xinyi Zeng, Zhaoxin Fan, Wei Ye — Thu, 05 Nov 2026 00:00:00 +0000

98% · Agents 98%, Evaluation 44%

Tool-using large language model (LLM) agents translate model outputs into consequential actions, turning indirect prompt injection into an execution-security problem. Existing defenses constrain tool calls, trajectories, or information flows, but often leave action semantics, state dependencies, and evidence relations implicit, limiting generalization across heterogeneous tools and complicating safe updates. We present \ours{}, a dynamic ontology-guided governance framework. \ours{} grounds heterogeneous calls in a shared semantic representation and combines an executable ontology with episode-level task state to derive deterministic ready, waiting, and forbidden decisions. A waiting decision identifies an unmet but recoverable obligation, enabling bounded replanning without expanding the agent's authorization boundary. \ours{} evolves its ontology offline from verified training executions. An offline LLM proposes instance-independent action--state axioms, which are accepted only if they address ontology errors exposed by training failures, preserve logical consistency and protected entailments, and pass held-out validation before the ontology is frozen. We evaluate \ours{} on AgentDojo, ASB, and AgentDyn under benign and indirect prompt-injection settings. \ours{} achieves a 0\% attack success rate (ASR) across all three benchmarks while improving task utility under attack compared with existing baselines. These results show that explicit semantic governance can strengthen execution security without indiscriminate blocking.

★ LifeVal: Evaluating AI Model Performance on Personalized Agentic Consumer Tasks

Thu, 05 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 53%

We introduce LifeVal, a benchmark evaluating AI model performance on personalized agentic consumer tasks. LifeVal covers the most valuable life tasks that a personalized digital superassistant could automate, representing over 1.7 trillion US dollars in annual willingness-to-pay across 148 consumer task categories (e.g., managing finances, moving apartments, searching for a job, finding childcare). To construct our activity distribution, we conducted a new survey of willingness-to-pay for a digital superassistant across a demographically representative sample of over 3,800 consumers, cross-referencing our results against the American Time Use Survey (Bureau of Labor Statistics, 2025) and the Multinational Time Use Study (Fisher et al., 2025). For the most valuable consumer activities, we recruited task creators to create AI evaluation tasks personalized using real ChatGPT conversations and memories, reference files, and personal context. Task creators had to pass our expertise threshold (including personally completing the task multiple times and seeing the results through in their real lives for over a year) and also write a rubric, ensuring that our evaluation measured the best personalized completion for that task. All tasks had to pass at least three rounds of human quality-control before acceptance into our final evaluation sets.

★ Reasoning over Non-Parametric Memory in Multilingual LLMs: A Cognitive-Based Analysis of Relational Knowledge

Zohreh Abbasi Liasi, Batoul Hammoud, Anas Madkoor, Ehsaneddin Asgari — Thu, 05 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

Retrieval-augmented generation (RAG) assumes that language models can reason over relations stated in their context rather than fall back on associations memorized during pretraining. We study this evidence-conditioned relational reasoning in isolation, with the facts supplied directly in the prompt rather than retrieved. \textsc{Rel-Bloom} is a controlled multilingual diagnostic of 450 multiple-choice questions built from 30 Wikidata relations; the same source relations are instantiated at five Bloom-inspired levels (Remember, Understand, Apply, Analyze, Evaluate) and under three veracity conditions (Factual, Counterfactual, Neutralized). Across 43 models, six languages, and two multilingual settings (226,800 predictions), three findings stand out. First, accuracy drops sharply from Apply to Analyze in every evaluated open-weight model series, reaching 47.2 percentage points for the Large Granite bucket, but not for either Gemini model. Second, Evaluate items remain a bottleneck even for models near ceiling on the other task types: Gemini~3.1 Pro averages 99.5\% on the other four levels but 60.7\% on Evaluate. Third, item- and relation-level difficulty rankings are consistently preserved across languages, indicating a difficulty structure partly shared across languages within this benchmark. Across knowledge conditions, Counterfactual contexts are generally harder than Factual ones, while removing entity identities can improve higher-level performance in some model regimes. Greater model capability therefore does not by itself ensure reliable reasoning over supplied relational evidence. Relevant resources are available at: https://huggingface.co/spaces/QCRI/non-parametric-memory-relational-reasoning.

★ ServerBench: How can your coding agents design specialized LLM serving systems?

Xinrui Zhong, Zhiming Wang, Yongqi Liu, Zhaohui Wang, Shanglin Yang, Zikun Li — Thu, 05 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Efficiency 72%

Large language model (LLM) coding agents develop software through experimentation, offering an opportunity to design the serving systems supporting their own operation. Effective serving requires coordinating implementation and operating decisions under session dependencies, latency constraints, and correctness requirements. We introduce ServerBench, a benchmark for agentic design of complete, specialized LLM serving systems. Agents experiment against a common objective, dollars per dollar: latency-qualified token value per dollar of GPU cost. Stateful replay preserves session dependencies, and independent verification checks the submitted system. Across 11 coding-agent models and three deployments, Claude Sonnet 5.5 achieves 2.97 dollars per dollar on Gemma E4B, $4.87\times$ the vLLM reference; 22 of 33 runs yield verified systems, and 13 exceed the strongest baseline. Our analysis identifies three findings: high-value designs coordinate capacity improvements with load selection; agents interpret performance feedback through workload-relevant experiments; and reliable delivery requires verification of the execution paths introduced by optimization. Overall, ServerBench supports systematic study of how agents translate service objectives and feedback into effective, verified serving systems for given task specifications.

★ FrontierPhysics: Evaluating Agents for End-to-End Physics Research

Bingran You, Wenjun Ke, Ria Rosenauer, Bo Liu, Wenbo Chen, Min Chen — Thu, 05 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 76%

AI agents are playing an increasingly important role in physics research. However, existing physics benchmarks only evaluate fixed questions, isolated sub-tasks, or paper reproduction with given solutions, and rarely check research quality alongside correctness. We introduce FrontierPhysics, an end-to-end benchmark with 56 tasks made by 42 physicists based on their own research, covering nine domains across theoretical, computational, and experimental physics. Each task requires agents to review the literature, develop a research plan, conduct research and experiments, and produce structured outputs and scientific manuscripts. Combining deterministic verifiers with expert-authored rubrics, FrontierPhysics evaluates whether agents' research is correct and publishable. Across 22 models in 26 configurations, no configuration passes more than 46 of its 168 rollouts. Manual review of agent trajectories shows that frontier agents' main weaknesses are insufficiently rigorous reasoning and limited expert intuition, rather than gaps in physics knowledge or execution errors.

★ BEAKER: An Expert-Curated Benchmark for Embodied Brains in Self-Driving Chemical Laboratories

Fei Lin, Tengchao Zhang, Ziyang Gong, Xiaotong Yu, Bohan Zhang, Yifan Zhou — Thu, 05 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

Self-Driving Chemical Laboratories (SDCLs) are moving chemical experimentation toward embodied automation, where machines must perceive laboratory scenes, reason about experimental states, and act under physical, procedural, and safety constraints. Although Multimodal Large Language Models (MLLMs) are increasingly considered as embodied brains for such systems, existing laboratory benchmarks mainly focus on safety, anomaly detection, or general scientific understanding, leaving their execution-oriented embodied capabilities underexplored. To fill this gap, we introduce BEAKER, a Benchmark for evaluating Embodied Actionability, Knowledge, Experimentation, and Reasoning in SDCLs. BEAKER contains 1,000 expert-curated question-answer samples across 11 image and video task types, covering key judgments required for chemical laboratory execution, including affordance grounding, trajectory planning, equipment, state, and safety recognition, apparatus setup, operation and step reasoning, and temporal procedure reasoning. All questions and annotations are manually created and reviewed by chemistry and embodied Artificial Intelligence (AI) experts, without AI-assisted generation or annotation. We evaluate 28 general-purpose, science-oriented, chemistry-specialized, and embodied MLLMs in an offline, single-turn setting. Results show that current models can follow coarse experimental procedures in videos, but still struggle to ground actions into operable regions and feasible trajectories, to recognize fine-grained laboratory apparatus, states, and safety norms, and to order steps in mid-length procedures. BEAKER thus reveals a clear gap between coarse procedural understanding and the precise, action-grounded perception required for laboratory execution, providing a diagnostic benchmark for future MLLMs in SDCLs.

★ Agent Harness Awareness: Turning Verified Telemetry into Operational Control for Tool-Using Agents

Manu Agrawal, Shrey Nagpal — Thu, 05 Nov 2026 00:00:00 +0000

97% · Agents 97%, Evaluation 93%, Efficiency 55%

Tool descriptions tell an agent what an action can do, but they do not necessarily reveal whether that action is healthy or suitable at the moment of use. Execution harnesses often possess this information through monitors, traces, and recent failures while the model choosing the next action does not. We call this gap operational blindness. We introduce Agent Harness Awareness, a trust-separated design in which the host supplies a small, fresh, verified state snapshot before a consequential action while retaining measurement, enforcement, evaluation, and final dispatch authority. We test its tools-and-network instance with ToolRoute, a controlled isolation benchmark that holds the task and candidate semantics constant and varies the operational information available to the model. In a 216-decision ablation across GPT-5.6 Sol, Claude Sonnet 5, and Gemini 3.7 Flash, verified tool-health state reduced mean observable policy regret by 96.9% against a same-shape neutral control (4,252 ms to 131 ms) and raised single-dispatch completion from 69.4% to 95.8%. In a separate 27-decision end-to-end replication with live local TCP/HTTP execution, OpenTelemetry client spans, and Toxiproxy-injected faults, verified state reduced mean regret from 8,956 ms to 0 ms (zero excess observable operational cost) and completed the live action in 9/9 model-by-fault cases versus 5/9 for the neutral control. These studies establish the tools-and-network instance of Agent Harness Awareness: selected host telemetry materially improves pre-dispatch operational choice. Runtime state, hardware pressure, economics, retry-aware recovery, and heterogeneous selection remain open empirical extensions of the same architecture.

★ Can Large Language Models Review the Scientific Literature? A Test Against a 102-Article Benchmark

Thu, 05 Nov 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 77%, Hallucination 69%

Large language models (LLMs) are increasingly used to support scientific literature reviews, yet evaluations often examine individual tasks rather than the review process as a whole. We assess six AI tools across four stages: finding relevant articles, interpreting them, synthesizing across them, and proposing future research. The benchmark is a published review of the Spanish plume weather pattern and its 102-article corpus. In the initial searches, five tools collectively identified 57 articles, with the best individual tool finding 42. Institutional library access increased discovery from 42 to 82 articles for Codex and from 15 to 79 for Claude Code. Given the complete corpus of 102 PDFs, three tools correctly identified all 18 articles discussing the Spanish plume substantially, but incorrectly classified 40% to 79% of the 84 passing mentions as substantial. The same three tools produced syntheses evaluated blindly by two domain experts. Although largely accurate, the syntheses provided limited insight and sometimes contained unsupported claims, quotation errors, and exaggerated statements of consensus. Finally, all three tools proposed research directions broader than the Spanish plume literature, although two approached questions identified in the benchmark review. No single tool performed best across all tasks. We provide eight recommendations for researchers using AI tools in literature reviews. Our findings suggest that LLMs can assist with discovering and interpreting scientific literature, particularly when institutional access is available, but cannot replace expert judgment in evaluating evidence or synthesizing research.

★ Cognitive Bias Signatures in Search-Equipped LLM Agents: A Stage-Resolved Analysis

Yanru Xiao, Yingwei Yu — Thu, 05 Nov 2026 00:00:00 +0000

97% · Agents 97%, Evaluation 92%

Search can help LLM agents correct mistakes, but their starting conditions may shape which evidence they find and use. Drawing on cognitive-bias paradigms, we develop a controlled search harness to examine result order, records of prior work, and initial hypotheses while holding each question and document collection fixed. Matched controls distinguish order effects from differences in document composition and the effects of failed work from context length. Across 10 models and 1,000 synthetic search worlds, we track queries, retrieved documents, document openings, and final answers. Our comparisons suggest that final answers are more sensitive to initial evidence order during active search than when all evidence is supplied in full. Dead-end histories can reduce strategy switching relative to matched neutral histories, indicating an effect beyond context length. Under wrong hypotheses, stage measurements distinguish reduced retrieval or opening from errors that persist after counter-evidence is opened. Exploratory interventions show that opening counter-evidence more often can coincide with lower accuracy, while accuracy can improve with little change in opening.

★ PAYWORTHY: Can LLMs Predict Which Freelance Jobs an AI Agent Will Complete?

Thu, 05 Nov 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 88%

Before an AI agent takes on a paid job, someone has to judge whether it is likely to succeed. We test whether frontier language models can make this call before any attempt. We propose PAYWORTHY which contains 1{,}494 real freelance job postings, sampled from 18.8M postings made between 2020 and 2026 on a freelancing platform. An agent attempts each job from the posting alone in a single run, and an LLM judge acting as the client decides whether to accept the work. We ask 18 models to forecast how likely the agent is to succeed and measure how well each model's forecasts rank the accepted jobs above the rejected ones (AUC, where 0.5 is a random ranking). The answer depends on how the agent is described. Told what the agent cannot do, every model ranks the jobs well (AUC 0.66--0.86). Told instead that it is a generic agent with a browser and code execution, the same models rank the same jobs about as well as a random ordering (0.45--0.56). These results hold under judges from four other model families (Claude, GLM, Kimi and Qwen), and the informed forecasts also rank the verdicts of two human raters. Code is available at https://github.com/upwork/Payworthy

★ ModalSearcher: Weaving Visual Evidence into Autonomous Cross-Modal Deep Search

Zhihan Yin, Ruyi Li, Chengyu Fang, Yifeng Yao, Nonghai Zhang, Pengfei Xiong — Thu, 05 Nov 2026 00:00:00 +0000

97% · Agents 97%, Evaluation 97%, New model 83%

Static and incomplete parametric knowledge limits the performance of multimodal large language models on novel-knowledge and knowledge-intensive tasks, especially those requiring multi-hop evidence acquisition. Existing multimodal agentic search methods often fix visual evidence at the input or a predefined trajectory position, making it difficult for agents to select search modalities as their information needs evolve and to handle complex real-world queries. We define this capability as autonomous cross-modal search: an agent chooses among direct reasoning, text-to-text search, text-to-image search, and reverse-image search, actively introduces visual evidence when appropriate, and switches between and integrates modalities across multiple hops. We propose ModalSearch-Pipeline, a scalable procedure for constructing an interleaved multimodal knowledge graph. It samples paths with diverse depths and modality compositions and uses backward question construction to obscure entity names and image targets that would expose reasoning shortcuts while retaining necessary disambiguating clues, yielding text-only, image-initialized, and arbitrarily interleaved multi-hop problems. We train the ModalSearcher Agent with supervised fine-tuning and reinforcement learning in a live Web environment, and construct ModalSearch-Bench. Experiments show that ModalSearcher-8B improves average accuracy by 16.2 points over Qwen3-VL-8B under the same agentic workflow across five multimodal knowledge-intensive QA benchmarks, with consistent gains on independently constructed active visual-search benchmarks.

★ HarnessFlow: MCTS-Guided Harness Optimization for LLM Agents

Zhi Hong, Qianlong Wang, Hongchen Wei, Aidan Wu, Daohong Ying, Mingze Kong — Thu, 05 Nov 2026 00:00:00 +0000

97% · Agents 97%, Efficiency 65%, Evaluation 63%

The performance of large language model (LLM) agents depends on both the underlying model and the harness that governs interaction with tools and the environment. Yet harness design remains largely manual, and automating it requires navigating an open-ended search space through costly end-to-end evaluations. We introduce HarnessFlow, a framework for optimizing complete executable harnesses under a limited candidate-evaluation budget, with the model and task environment held fixed. HarnessFlow organizes candidates and their evaluation histories in a refinement tree, using a Monte Carlo Tree Search (MCTS)-based policy to balance exploitation and exploration. A coding agent refines selected harnesses using archived code, execution feedback, and a library of reusable harness mechanisms. With 15 refinement attempts per benchmark, HarnessFlow achieves the highest mean test scores among the compared methods on six benchmarks spanning agentic tasks, coding, and web interaction. The optimized harnesses achieve an average score of 69.48%, outperforming the initial harnesses and Meta-Harness by 10.89 and 3.95 percentage points, respectively, while incurring lower average per-task execution costs than hand-engineered baselines. These results suggest that structured harness search offers a promising route toward automating the design of capable and efficient LLM agents.

★ Task-Efficient Agent Evaluation Using Cached Traces

Ben Johnson, Hayden Helm — Thu, 05 Nov 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 57%, Efficiency 53%

Efficient benchmarking estimates a system's full benchmark performance from a small number of task executions. Standard approaches rely solely on the binary correctness of executions from a collection of already-evaluated reference systems. An agentic system, however, produces a rich trace for each execution that can reveal substantially more about its capabilities. Leveraging this information requires useful trace representations. We propose representing a trace by extracting a structured summary of it with an LLM and embedding the summary with a standard text encoder. These representations preserve correctness- and task-related signal while minimizing the presence of system identity-, model family- and harness-related signals that dominate raw trace embeddings. Given this representation for each task execution from all reference systems, we construct low-dimensional representations of the systems and predict a new system's score from its position among the reference systems. On SWE-bench Verified and Terminal-Bench 2.0, augmenting an item-response model with this geometry-based prediction reduces mean absolute error from 0.13 to 0.08 (Wilcoxon $p < 10^{-4}$) and from 0.14 to 0.12 (Wilcoxon $p < 10^{-3}$), respectively, when provided a single random executed task. Further, given access to the full set of reference systems, our approach dominates baselines for any given number of executed tasks on both benchmarks.

★ SAUCE: Summary Analysis Using Conversation Entailment

Man-Ling Sung, Hemanth Kandula, Jeff Ma, William Hartmann, Matthew Snover — Thu, 05 Nov 2026 00:00:00 +0000

97% · Evaluation 97%, Hallucination 95%

With the growing need for evaluating Large Language Models (LLMs) and their applications to speech, challenges persist in summarizing and evaluating conversations that lack a clear end goal. We introduce SAUCE -- a reference-free, fact-based evaluation pipeline for cross-lingual conversational speech summarization. It measures the accuracy and the fact coverage of a summary through the entailment between conversation and text. We compare SAUCE against several popular summarization metrics and demonstrate the effectiveness of capturing information loss due to transcription and translation error and identifying broken summaries. Crucially, unlike black-box LLM evaluators or dense embedding metrics, SAUCE is inherently explainable: it maps summary scores to discrete, verifiable facts, allowing users to pinpoint exact hallucinations or omissions. We illustrate how this interpretability helps developers systematically profile LLM behaviors and gives end-users an actionable tool to verify summary accuracy in noisy, real-world conditions. Preliminary investigations show SAUCE strongly align with human judgment.

★ Hybrid Spark: Adaptive Speculative Decoding That Lets the Content Choose the Drafter

Ekagra Ranjan, Bharat Venkitesh — Thu, 05 Nov 2026 00:00:00 +0000

96% · Efficiency 96%

Speculative decoding (SD) speeds up large language model (LLM) serving, but its gain depends on the workload's \emph{shape}, such as batch size and context length, and on its \emph{content}. A model-based drafter proposes a short block on any content, while a model-free drafter such as n-gram lookup proposes long copies at almost no cost when the output repeats its input, as in code editing. Adaptive verification spends a batch-wide verification budget on the draft tokens most likely to be accepted, but it supports only one model-based drafter. Model-free drafters provide no per-token confidence, so they cannot take part in budget allocation, and deployments must either fix one drafter or split their GPUs between drafters. We present \sys, a framework that runs a model-based and a model-free drafter on every step and gives each request the better of their chains, within adaptive verification and entirely on the GPU, which keeps asynchronous scheduling intact. The key idea is a price per verified draft token: the goodput that slot would earn elsewhere in the batch. Once the model-free drafter's tokens carry calibrated acceptance estimates, fit once on general text, this price puts both drafters' chains on one scale. Instantiated with \dspark and an n-gram drafter in vLLM, \sys achieves 1.10--1.62\x the output throughput of \dspark on code editing at batch size~4 and 1.04--1.27\x at batch size~128, while staying at 0.97--1.01\x of it on chat at temperature~0. A single deployment can therefore serve both workloads.

★ RECAP: Context-Aware Prediction Recycling with Request-Local and Persistent Memories for Multimodal Speculative Decoding

Hengjie Zhu, Dayan Wu, Xinze Liu, Peng Fu, Zheng Lin, Weiping Wang — Thu, 05 Nov 2026 00:00:00 +0000

96% · Efficiency 96%, Evaluation 40%

Recycling a target model's own predictions offers a training-free source of proposals for speculative decoding. In vision-language models, however, existing recycling indexes predictions by the last token alone, conflating predictions computed under different images and prefixes; we find that such reuse fails most often at image-sensitive decoding states. Longer keys reduce this aliasing but match fewer states. We introduce \method, which stores target predictions under one-, two-, and three-token keys: contextual memories live only for the current request and are thus implicitly conditioned on its image, while token-level memory persists across requests to retain coverage. At each lookup, entropy-confidence weighting fuses the retrieved predictions according to their key specificity and concentration. Best-first search composes the fused transitions into a candidate tree, supplemented by suffix-matched continuations, and one target pass verifies the tree under greedy decoding. Without training or an auxiliary draft model, \method\ achieves mean speedups over autoregressive decoding of $1.66\speed$ and $1.56\speed$ across seven single-turn benchmarks on Qwen2.5-VL-7B and LLaVA-1.6-7B, exceeding the strongest training-free baseline, Token Recycling ($1.51\speed$ and $1.37\speed$). Our implementation is available in https://github.com/Senmo996/RECAP-official-code.

★ Overview of the NTCIR-19 SciClaimEval Task: Cross-modal Claim Verification in Scientific Papers

Thu, 05 Nov 2026 00:00:00 +0000

96% · Evaluation 96%

This paper presents an overview of \emph{SciClaimEval}, a pilot task at NTCIR-19 on verifying scientific claims against the tables and figures of the papers in which they appear. Our dataset pairs authentic claims from 180 papers in machine learning, natural language processing, and medicine with contrastive versions of their tables and figures, one supporting the claim and one minimally modified so that it does not. The task comprises two subtasks: deciding whether a piece of evidence supports a claim, and selecting which of two evidence versions does. Twenty-two teams registered, the highest number among NTCIR-19 tasks, and 14 submitted 177 runs. Most systems built on multimodal large language models, relying on prompting and multi-stage inference rather than task-specific fine-tuning. On the first subtask, the best systems reached 98.1\% pair accuracy, but all six top-ranked systems relied on the contrastive pairs; the best system that judged each claim-evidence pair on its own scored 24.3 points lower. Open-weight models, embedded in carefully designed pipelines, matched proprietary ones. Yet verifying a single claim against its evidence, the setting that matters in practice, remains an open problem.

★ Beyond Risk Cues: A Mechanism-Level Benchmark and Verifiable Debate for Scam Video Detection on the Web

Yifan Wang, Junyu Lu, Shun Zhang, Chaozhuo Li, Fanliang Bu, Rong Yin — Thu, 05 Nov 2026 00:00:00 +0000

96% · Evaluation 96%, Hallucination 50%, New model 34%

Online scam videos pose significant threats to individuals and society. Existing studies primarily focus on binary classification, offering limited insights into how scams work and the reasoning behind detection decisions. Moreover, they often overlook the challenge of distinguishing scam videos from superficially similar non-scam content, such as advertisements, and the robustness of detection models against misleading arguments. These limitations hinder reliable scam video detection. To address these limitations, we introduce MechScam, a benchmark comprising 6,614 videos covering diverse scam mechanisms and non-scam categories. MechScam provides explanations supported by video evidence, annotations of potential scam elements and their context, and misleading arguments constructed from genuine evidence. Recognizing that genuine evidence can still be used to justify misleading conclusions, we further propose ScamCourt, a debate framework in which trained prosecution and defense debaters present opposing arguments, while a trained judge makes decisions based on a transcript with verified citations. Extensive experiments show that state-of-the-art MLLMs struggle to distinguish scam videos from similar non-scam content. ScamCourt outperforms larger MLLMs and same-backbone fine-tuned baselines, reducing false positives and showing stronger resistance to misleading arguments. Together, MechScam and ScamCourt contribute to more reliable and interpretable scam video detection. The dataset and code are available at https://github.com/SuperYFan/MechScam.

★ AssameseLM-E2B: An Open-Weight Assamese Language Model for Efficient Consumer-Grade Deployment

Pragyan Phukan — Thu, 05 Nov 2026 00:00:00 +0000

96% · New model 96%, Efficiency 85%, Hallucination 79%

While Large Language Models (LLMs) excel in high-resource languages, their performance often degrades in low-resource languages due to tokenization inefficiencies, generative instability, and linguistic interference. This paper presents AssameseLM-E2B and its instruction-tuned variant, AssameseLM-E2B-it, a 2-billion-parameter model adapted specifically for low resource Assamese language. Built upon the Gemma-4 architecture, the model incorporates an expanded byte-level BPE vocabulary to reduce tokenization overhead by nearly 500% compared with multilingual 2B--3B baselines. We employ a parameter-efficient Continual Pre-Training (CPT) pipeline with Hybrid LoRA, followed by Supervised Fine-Tuning (SFT) on a balanced, multi-domain dataset to enhance instruction-following and Retrieval-Augmented Generation (RAG) capabilities. Evaluation demonstrates a Bits-per-Byte (BPB) score of 0.7488 on FLORES-200, alongside an overall RAG score of 96.7% and 100% refusal accuracy on adversarial unanswerable prompts. Comparative analysis reveals significant limitations in general-purpose baselines, including receptive-productive decoupling, generative degradation on Romanized Assamese, and unintended Bengali script assimilation. These results demonstrate the potential of parameter-efficient adaptation for developing capable, locally deployable language models for low-resource languages under constrained computational resources.

★ D3D: Decoupled Diffusion Drafting for Asynchronous Speculative Decoding

Baoyou Chen, Ruiqiao Mei, Guancong Lin, Tao Song, Weihao Yuan, Qun Liu — Thu, 05 Nov 2026 00:00:00 +0000

96% · Efficiency 96%, New model 69%, Evaluation 31%

Target-conditioned diffusion drafters predict candidate tokens in parallel, yet each new block must wait for updated target-model hidden states. This forces drafting and verification to run sequentially, limiting speedup even though each block is generated in one pass. We introduce D3D, a standalone diffusion drafter that removes this inter-round dependency. D3D combines blockwise denoising with auxiliary autoregressive supervision to achieve long accepted prefixes without target hidden states. This independence lets D3D draft successor blocks during verification in one packed pass, enabling asynchronous speculative decoding. We progressively prune layers according to their effect on acceptance length to reduce the cost of packed drafting. Removing five of 36 layers retains over 99\% of mean acceptance length across three benchmarks with Qwen3-8B as the target. Across seven benchmarks, D3D averages 5.00$\times$ and 5.71$\times$ decoding speedup with Qwen3-8B and Qwen3-32B targets, compared with 4.03$\times$ and 3.67$\times$ for DFlash. On GSM8K with Qwen3-32B, speedup reaches 7.48$\times$. Code is publicly available at https://github.com/fudan-generative-vision/D3D.

★ SUMONE: A Simple and Efficient Summarizer for Plug-and-Play Agent Memory

Jizhe Zhang, Ang Li, Chenheng Zhang, Yisen Wang — Thu, 05 Nov 2026 00:00:00 +0000

96% · Efficiency 96%, New model 87%, Agents 75%

Long-horizon language-model agents accumulate increasingly large interaction histories, making context compaction essential for efficient execution. Existing learned compaction methods often rely on downstream task outcomes or solver behavior, coupling training to specific tasks or agents. We introduce SumOne, a standalone summarizer trained to preserve information in agent histories through masked reconstruction. Given a summary of the expired history, a frozen reconstructor recovers randomly masked portions of the original trajectory; reconstruction quality, combined with a length penalty, provides a task-agnostic reward for optimizing the summarizer with GRPO. This yields self-supervision directly from raw trajectories, without reference summaries or downstream task execution. At inference time, only the lightweight summarizer is retained, enabling plug-and-play context compression for otherwise unchanged agents. We conduct a systematic evaluation of SumOne in long-horizon search and software-engineering tasks. SumOne reduces estimated inference cost by approximately 78\% on DeepSearchQA and 75\% on SWE-bench Verified relative to operating with full context, with absolute decreases of 0.04. The same checkpoint also transfers across downstream models and agent harnesses without additional fine-tuning.

★ Lost in Recovery: Demystifying Recovery Behavior of Computer-Use Agents via Goal-Oriented Action Replay

Thu, 05 Nov 2026 00:00:00 +0000

96% · Evaluation 96%, Agents 92%

Recent advances in large (vision) language models have enabled computer-use agents that automate digital workflows via fine-grained keyboard and mouse operations. However, such agents remain fragile in long-horizon settings, where errors accumulate and executed actions leave unintended side effects, making the ability to recover from failures critical. Yet, existing remedies remain inadequate: environment resets are infeasible in proprietary settings, action replay is prohibitively expensive over long horizons, and API-based rollback is limited to web environments, leaving model-based recovery largely unexplored. To address this gap, we present ReTrace, a scalable evaluation framework for systematically studying recovery behavior in computer-use agents. We formalize Transition-Consistent Recovery: identical downstream actions from the recovered and target states must yield identical outcomes. We operationalize this criterion through goal-oriented action replay following controlled error injection. Using ReTrace, we present the first systematic study of autonomous, model-based recovery, revealing that (i) ideal recovery can unlock model's latent task completion capabilities, (ii) current computer-use agents fail to exploit their latent recovery capabilities, and (iii) accurate model-based recovery yields measurable end-to-end gains. Together, these findings establish recovery as a fundamental capability for building reliable computer-use agents. Our code and dataset is available at https://github.com/BangProx/ReTRACE.

★ C-GRPO: Conformal Group Relative Policy Optimization

Arya Fayyazi, Seyedarmin Azizi, Mehdi Kamal, Massoud Pedram — Thu, 05 Nov 2026 00:00:00 +0000

95% · New model 95%, Efficiency 82%, Evaluation 66%

Reasoning-oriented language model training increasingly relies on multiple on-policy rollouts, but standard Group Relative Policy Optimization (GRPO) uses a fixed group size, wasting compute on easy prompts and under-exploring hard ones. Existing adaptive-budget methods alleviate this inefficiency, yet their stopping rules are heuristic and do not provide a finite-sample calibration guarantee. We propose CGRPO, which replaces the fixed rollout budget of standard GPRO with calibrated conformal thresholds over task-specific nonconformity scores, combining Adaptive Prediction Set (APS)-style calibration for closed-form reasoning, a first-success score for execution-based code, automatic miscoverage selection, and periodic re-calibration. We prove finite-sample coverage for the resulting stopping rule and show that, under monotone policy improvement, recalibration induces a non-increasing expected training budget. Across six benchmarks spanning mathematics, graduate-level science, and code generation, CGRPO reduces mean training rollouts by 49-70% relative to fixed-budget GRPO while retaining strong fixed-budget accuracy, achieves the strongest pass@1 among evaluated static and dynamic baselines on different datasets, and transfers across model families with up to 82% rollout savings.

★ Information-Matched Auditing of Conversational Emotion-Shift Forecasting: A Leakage-Safe Evaluation Protocol

Bin Wen, Dai-Qiao Zhang, Tien-Ping Tan — Thu, 05 Nov 2026 00:00:00 +0000

95% · Evaluation 95%

Comparisons in conversational emotion-shift forecasting are easily distorted when a learned model receives rich features but its transition baseline is given a noisy or unavailable current-emotion label, when hyperparameter budgets differ, or when target-turn information enters the representation. We present an information-matched, leakage-safe auditing protocol that closes these gaps with an explicit commitment point, executable causality tests, equal four-configuration validation searches, and seed-dialogue-aware inference. Applying it to six causal model families on four text corpora and two feature-source sensitivity configurations shows that a deployable train-only current-emotion classifier feeding the transition baseline yields a different verdict than an oracle diagnostic where both sides receive gold current emotion. Under the deployable condition, learned models improve over the predicted-label baseline on all four primary corpora; DailyDialog gains are large and survive a 72-comparison Holm correction (best ROC-AUC gain = +0.167, adjusted p=.0144). Under matched gold information, the baseline remains competitive on IEMOCAP and stronger on MELD, while every model retains a large DailyDialog advantage. We additionally report precision-recall AUC, calibration, a strictly utterance-local feature control, same-person self-shift forecasting, and a future-access dose-response test. An eight-item leakage checklist accompanies the protocol. Code, raw seed predictions, and reporting artifacts are available from the corresponding author upon request.

★ Routes of Reasoning: Reading and Reshaping Test-Time Compute in MoE Models

Shengjie Wang, Mohsen Fayyaz, Nanyun Peng — Thu, 05 Nov 2026 00:00:00 +0000

95% · Efficiency 95%, Evaluation 55%

Reasoning models allocate test-time compute across distinct behaviors such as planning, deriving, and verifying. However, models frequently "overthink," continuing to verify and re-derive long after finding a correct answer. This wastes expensive compute and can even degrade performance. Existing interventions, such as token budgets or steering, primarily regulate reasoning duration rather than targeting specific behaviors. In this work, we demonstrate that Mixture-of-Experts (MoE) routers provide a natural, training-free interface for behavioral control. We first split reasoning traces into sentence-level segments and label each with one of six reasoning behaviors (planning, execution, restatement, fact, reflection, and conclusion), so that every token carries the behavior of its segment. Linear probes on router logits predict this label nearly as well as probes on the residual stream, the model's full hidden state and the standard target for probing, while using about half as many features. These signals generalize: experts associated with a reasoning behavior on one dataset predict the same behavior on the others, across six datasets spanning five domains (math, code, science, knowledge, and commonsense). Building on this, we selectively edit expert router scores during inference to curb overthinking. Across four models and four benchmarks, this intervention reduces the number of generated tokens by a median of 30.4% (and up to 63.3%). Crucially, it encourages the model to stop once the problem is solved, rather than skipping essential reasoning steps. Accuracy never drops by more than 0.5 percentage points, and frequently improves. Ultimately, MoE routing offers both an interpretable lens into reasoning states and a lightweight mechanism for optimizing test-time compute.

★ ConceptARC at Scale: Studying Conceptual Understanding in Reasoning Models

Thu, 05 Nov 2026 00:00:00 +0000

95% · Evaluation 95%

Large reasoning models (LRMs) perform well on challenging benchmarks, yet it remains unclear whether they form generalizable abstractions across conceptually related reasoning tasks. The Abstraction and Reasoning Corpus (ARC) provides a controlled testbed for studying abstraction, requiring solvers to infer transformation rules over two-dimensional grids from a small set of examples. While ARC evaluates rule abstraction within individual tasks, different tasks may share higher-level structure despite requiring different transformations$\unicode{x2014}$for example, several tasks may involve reasoning about objects contained within other objects. Benchmarks such as ConceptARC capture this structure by grouping ARC-style tasks into recurring spatial and semantic concepts, but they often contain only a few manually constructed tasks per concept. We introduce ConceptARC at Scale, a generator for creating thousands of ARC-style tasks targeting a specific concept. Using this generator, we construct ConceptARC-100, a dataset of 1,600 new, manually reviewed tasks, with 100 tasks for each of the 16 concepts defined by ConceptARC$\unicode{x2014}$ten times the size of the original benchmark. We evaluate a range of recent LRMs and models trained from scratch on ConceptARC-100 to characterize which concepts they solve reliably and where they struggle. We further analyze the models' internal representations and find clear semantic structure: task representations cluster according to the operations applied to objects, such as recoloring, removal, or translation, even when realized through substantially different rules and scenarios. Together, our results suggest that models form stable abstractions of transformation operations that generalize across diverse tasks.

★ CompressionBench: How can your coding agents compress LLMs?

Xinrui Zhong, Zhifeng Li, Zhiming Wang, Yongqi Liu, Shanglin Yang, Yang Zhou — Thu, 05 Nov 2026 00:00:00 +0000

95% · Evaluation 95%, Efficiency 66%

Recursive self-improvement requires AI systems to improve the models and infrastructure used by subsequent systems. LLM compression offers a concrete efficiency objective: retain a pretrained model's behavior within a smaller model storage budget. We study whether coding agents can carry out this step autonomously. We formulate compression as a terminal task in which an agent receives a source checkpoint, storage limit, inference environment, and time budget, then constructs compressed weights and executable inference code. The agent controls its algorithm and experiments; perplexity feedback guides search, and independent verification checks the final model. Across six agents and six model--size settings, 29 of 36 submissions pass verification. Selecting the lowest-perplexity agent design per task yields an average 3.4% perplexity increase over BF16 and 77.33% downstream accuracy, compared with 79.56% for BF16. Recorded experiments reveal how agents achieve these outcomes and where improvement breaks down. Agents lower PPL in 21 of 24 revisions at nearly fixed storage; complete-model feedback reverses 17 of 79 prior local rankings; and generation and inference tests expose failures that likelihood alone misses. The results identify representation design, experimental judgment, and capability validation as central to agent-driven model compression, a measurable component of efficiency-oriented self-improvement.

★ SpecCompact: Accelerating LLM Agent Context Compaction with Slot-Parallel Semantic Speculative Verification

Yulu An, Yifei Feng, Xiangyang Yin — Thu, 05 Nov 2026 00:00:00 +0000

95% · Efficiency 95%, Agents 56%, Evaluation 54%

LLM agents use context compaction to manage growing interaction histories, but summarizing long histories with large models can introduce substantial delays. Compaction therefore requires balancing information retention against latency. We observe that structured summary slots can retain information when generated independently. Based on this observation, we present SpecCompact, a framework that accelerates compaction through slot-parallel semantic speculation. SpecCompact fine-tunes a small drafter and generates slots in parallel. The target model verifies each proposed line and repairs rejected content, while subsequent drafting uses the updated slot prefix. System optimizations further reduce repeated context processing and model calls. Across AppWorld and $\tau^3$ benchmarks with two target models, SpecCompact achieves $2.04$--$5.63\times$ compaction speedups over Hermes-style summarization. Separate downstream evaluations reveal model- and task-dependent performance trade-offs. A user study within Hermes further shows an $8.68\times$ median compaction speedup.

VirtualTrialHarness: Coordinating Agentic Operationalization of Virtual Trial Protocols Across Complex Clinical Schemas

Pengze Li, Jingna Feng, Yue Yu, Yining Wu, Xinyue Hu, Jie Lin — Thu, 05 Nov 2026 00:00:00 +0000

94% · Evaluation 94%, Agents 60%

Virtual trials use routinely collected clinical data to evaluate treatment strategies, but translating a study protocol into an executable analysis requires determining how clinical concepts and study requirements should be represented in the available data and applied consistently across the workflow. These decisions depend on local data representations and operational conventions and can involve source selection, patient and encounter identity, linkage, and temporal interpretation. Here we introduce VirtualTrialHarness (VTHarness), an agent framework that coordinates virtual-trial operationalization by selecting reusable procedural skills for each workflow stage and uses explicit contracts to maintain consistent study decisions across stages. We also introduce VTBenchmark for evaluating source selection and cohort construction. Across three complex synthetic environments, VTHarness successfully produced all required cohort outputs matching the references in 95.1% of workflows after requirement clarification, compared with 60.4% using retrieval from the same procedural corpus. Procedural skills improved source recovery across three language models and increased source overlap with analyst code in 13 Mayo Clinic cases. Site-specific adaptation enabled reference-matched cohort execution across eight MIMIC-IV tasks. Within the Mayo Clinic Virtual Trials Program, the framework supported investigator review of cohort construction and endpoint analysis. These findings support coordinating procedural guidance with workflow decisions to improve virtual-trial implementation across complex clinical schemas.

Give the Drafter a Second Chance to Guess: Parallel Reconsideration for Speculative Decoding

Thu, 05 Nov 2026 00:00:00 +0000

94% · Efficiency 94%, New model 62%

Speculative decoding accelerates large language model inference with a small drafter, whose per-guess acceptance rate determines the achievable speedup. Feature-level drafters share a design choice: every token, easy or hard, is drafted under a single forward pass. Many of the tokens the target rejects are ones the drafter is itself unsure about yet the target answers decisively. To address this, we introduce Parallel Reconsideration: guided by its own uncertainty, the drafter reconsiders its hesitant guesses with extra passes that run in parallel with the one-pass draft, adding negligible latency. A co-designed training recipe and serving-time drafting algorithm append these reconsidered guesses to the original draft tree as refined candidates. Across models and benchmarks, on average, Parallel Reconsideration improves over the baselines by 5.3% to 10.2% on acceptance length and by 3.5% to 9.2% on speedup.

Addition by Subtraction: Toward Mitigating Fragile Safety Alignment in Large Language Models

Hamid Dashtbani, Alexandra Dmitrienko — Thu, 05 Nov 2026 00:00:00 +0000

94% · New model 94%, Hallucination 49%, Evaluation 36%

Large language models (LLMs) undergo safety alignment to prevent harmful generations, yet they remain vulnerable to jailbreak attacks. Recent work reveals a deeper structural weakness: safety alignment can be highly sparse, such that pruning a small number of safety-critical neurons or ablating a single refusal direction can substantially undermine aligned behavior. In this work, we take the first steps toward strengthening safety alignment such that it is less dependent on sparse model components. We propose Addition by Subtraction, which gradually removes identified safety-critical neurons during finetuning and uses preference optimization to re-establish safety-aligned behavior through other parts of the model. Our results show that the resulting models are substantially more robust to pruning-based attacks. Remarkably, these benefits generalize beyond the white-box attacks motivating our design: Addition by Subtraction improves robustness against other sparsity-based interventions, including refusal-direction ablation, as well as a broad range of black-box jailbreak attacks. The finetuned models also exhibit improved truthfulness while largely preserving general capabilities. Comprehensive experiments across four models from three vendors demonstrate consistent improvements in safety robustness, reducing attack success rates by up to 90% for some attacks.

Where Search Agents Go Off the Rails: Graph-Grounded Step-Level Verification

Hyeseo Jeon, Seonho Yoo, Jay-Yoon Lee — Thu, 05 Nov 2026 00:00:00 +0000

94% · Evaluation 94%, Agents 78%, Hallucination 75%

Search-augmented reasoning systems can reach correct answers while making errors in intermediate steps. Reliable auditing must check whether evidence $D$ covers the question $Q$, reasoning $T$ addresses $Q$, and $T$ is grounded in $D$. We introduce GraphRail, which represents requirements, evidence, and reasoning as structured claims and compares them without a generative judge or gold answer. Across three multi-hop QA benchmarks and three reasoning frameworks, GraphRail improves error recall over generative verifiers while remaining stable under stochastic extraction. Its diagnoses can also guide online reasoning, providing a traceable way to detect and correct intermediate errors.

Iterated creation, grading, and revision of data science projects by language models

Rohan Alexander — Thu, 05 Nov 2026 00:00:00 +0000

93% · Evaluation 93%, Agents 90%

We evaluate the extent to which agentic AI models can complete an end-to-end data science project, and how this improves with feedback. A 'writer' model with specific tools creates a complete data science project including downloading, cleaning, and analyzing a dataset, and writing a short paper about it. The project is executed and a 'grader' model evaluates it. The writer can then revise its submission, and this repeats for a total of three rounds. We consider seven models from five developers: Anthropic, OpenAI, Google, DeepSeek, and Z.AI. Each model is both a writer and a grader, so there are 49 writer-grader pairs and we run each pair three times. Of the 147 initial submissions, 107 run and render, and the average score increases from 59 per cent to 76 per cent after the final round of feedback. By writer the extent of the increase ranges from 3 percentage points for Anthropic's Opus 5.5 to 46 percentage points for Anthropic's Haiku 4.5, and by grader it ranges from 13 percentage points for Google's Gemini 3.8 Flash to 32 percentage points for Anthropic's Haiku 4.5. When a model grades its own writing the score tends to be about 9 percentage points higher than the average of what the other models give it. About half of the difference that we see from round to round is within the grader's own noise. Feedback generally results in the highest marks, but even just having the chance to revise, or having the additional budget, in terms of time or tool calls, without any feedback, results in considerable increases for some of the models but not for others. We find that how much such a loop helps depends on which model plays which role, but weaker models generally do not improve stronger models.

Rubric Curriculum RL: Exploiting the Generation-Verification Gap in Non-Verifiable Domains

Thu, 05 Nov 2026 00:00:00 +0000

93% · New model 93%, Evaluation 52%

Reinforcement learning with verifiable rewards (RLVR) has led to significant gains in math and code generation. Extending these gains to open-ended domains remains challenging: ground-truth verification is unavailable, human annotation is expensive, and learned reward models are prone to reward hacking. The generation-verification gap, where judging good outputs is easier than producing them, offers a potential path to improvement. However, rubric-based judges are often imprecise and stationary, so policies can overfit or reward hack them. We introduce Rubric Curriculum RL (RcRL), which exploits this gap without new data, human annotations, or stronger teacher models. We combine pairwise preferences over rollouts with a curriculum over rubric criteria, yielding a more discriminative signal than absolute scoring and a non-stationary objective. Whereas baselines plateau within a few hundred steps, RcRL preserves output entropy and keeps improving past 1000 steps. On creative writing, RcRL achieves a 70.5% human win rate against the base model, with gains across benchmarks and judges. On HealthBench, RcRL improves over the base model by 14.6% (full set) and 25.4% (consensus subset), outperforming all training baselines, including instance-specific rubrics.

WebHarbor: Docking Real Websites for Evolving GUI Agent Environments

Zhaoyang Wang, Qianhui Wu, Boyu Gou, Demi Wang, Jiacheng Zhang, Peichun Hua — Thu, 05 Nov 2026 00:00:00 +0000

93% · Agents 93%, Evaluation 88%

GUI web agents require realistic and interactive environments that remain stable during both evaluation and training. Live website evaluation suffers from changing content, access restrictions, and limited coverage of authenticated workflows, while existing synthetic environments for training often lack functional depth and visual fidelity. We propose WebHarbor, a suite of 153 executable and dockerized website environments that faithfully reconstruct the visual interfaces, content, and functional workflows of real-world web services under deterministic control. We build WebHarbor through task-driven human-agent collaboration, where environments evolve as new tasks expand their functional scope, coding agents implement the required changes, and human contributors audit visual fidelity, functional correctness, and task validity. The resulting suite comprises 3,518 tasks spanning multi-step and post-authentication workflows, where existing works often fail to cover. Beyond reliable evaluation, WebHarbor provides high-fidelity synthetic web environments for cost-efficient reinforcement learning (RL). We demonstrate that a Qwen 4B-level GUI agent trained with RL in WebHarbor generalizes to unseen tasks and performs competitively with leading open-source GUI web agents across four benchmarks. The code and data are publicly available at https://github.com/aiming-lab/WebHarbor.

Who May Overrule the Agent? Evidence-Gated Authority in LLM Review Institutions

Fuad Ali — Thu, 05 Nov 2026 00:00:00 +0000

92% · Evaluation 92%

Multi-agent language-model systems often add a reviewer to improve an upstream agent's decision. Yet a reviewer can both repair errors and replace decisions that were already correct. We study this as a problem of institutional jurisdiction: conditional on the same proposal, task information, model, and review output, when is the reviewer authorized to make its recommendation binding? We compare broad override authority with an evidence gate that permits replacement only after the reviewer produces a mechanically valid record of a binding-constraint violation and a compliant repair. Across two locally run open-weight models, 96 spatial-policy tasks, and a 96-task portfolio-transfer benchmark, the gate sharply reduces corruption of correct proposals while retaining most repairs of violations. The directions of this tradeoff are partly implied by the rule; its magnitude depends on reviewer transitions and upstream proposal states. A prospective validation then gives the reviewer an explicit choice to retain, replace, or escalate. This eliminates correct-proposal corruption for Qwen, but Mistral still replaces all 48 correct proposals; Qwen also declines every welfare-only repair that Mistral frequently attempts. Thus interface design changes reviewer behavior but is not a model-independent safeguard. The result is not that restricted review is universally better. It is that reviewer competence, action interface, and binding jurisdiction jointly determine both the correction domain and the corruption surface.

Route to Remember: Adaptive Memory Selection for LLM Agents

Zeyu Fang, Jun Yin, Yongsheng Mei, Xu Han, Chaitali Joshi, Zhongkai Sun — Thu, 05 Nov 2026 00:00:00 +0000

92% · Efficiency 92%, New model 83%, Evaluation 70%

The memory of production agents is increasingly becoming a bottleneck and is built almost exclusively from text: key points are extracted from out-of-limit context and retrieved into context at query time. Therefore, every new query incurs an additional cost that grows with the accumulated history. Weight memory is a complementary alternative that internalizes knowledge into model parameters and answers with no context at all, and has recently proven effective in self-evolving agents for internalizing skills and behaviors. However, its application in episodic and semantic memory, such as user-specific facts, remains comparatively unexplored, and a gap remains between earlier attempts at weight memory and text-based systems regarding recall performance. We study how to make weight memory competitive as a long-term factual store by applying on-policy self-distillation (OPSD). It trains a context-free student from a teacher, which is the same model conditioned on the user's memory, together with a verify-then-regenerate router that lets weight and text memory each cover the queries they handle best. On the HaluMem benchmark, OPSD matches or surpasses strong text-memory baselines, including full-context prompting and retrieval systems, while reading fewer tokens per query. We further conduct an extensive study that characterizes the behavior of weight memory and analyzes how different training recipes affect its performance. As agent harnesses move toward persistent, multi-session deployments, these results position weight memory as a practical, low-cost component of agent memory and provide insights into feasible training and evaluation protocols for combining it with retrieval.

Grounded but Not Immune: Confirmation Bias in Agent-as-a-judge

Thu, 05 Nov 2026 00:00:00 +0000

92% · Agents 92%, Evaluation 74%, Hallucination 52%

Evaluating open-ended agent tasks increasingly falls to an Agent-as-a-Judge: an evaluator handed tools and the agent's full trajectory so it can verify a result against the environment. The premise is that a judge that gathers its own evidence should be grounded by it. We show that this grounding does not immunize the judge against confirmation bias: a single false pre-screening note, attributed to an authority and carrying no evidential bearing on the correct verdict, overturns up to 40% of a tool-using judge's correct verdicts across four judges and five datasets, even though the judge collects decisive evidence. The shift is driven by specific properties of the cue. We treat this cue as a controlled diagnostic of confirmation bias rather than an adversarial attack. Our central contribution is to explain how confirmation bias changes the judge's behavior in a trajectory. We classify changes in judge behavior and attribute changes to one of three failure modes: over-trust (verify, conclude correctly, then defer), hallucination (fabricate support for the flip), and under-investigation (stop at the first corroborating signal). For the two Claude judges, the failure is predominantly over-trust (67-75% of harmful flips), while smaller judges have error modes spread across all three. Lightweight mitigations blunt but never remove the bias.

Shallow Encoding Emits Deep Memories

Mingyuan Ma, Yangmin Li, Harris Nover, Zheyu Fu, Po-Han Huang, Alex Settle — Thu, 05 Nov 2026 00:00:00 +0000

92% · Efficiency 92%, Evaluation 38%

Hybrid language models interleave a few softmax-attention layers with many linear-attention or state-space layers, so every layer keeps a memory of earlier tokens: a key-value cache or a fixed-size recurrent state. Serving them spends more compute and high-bandwidth memory on these memories than the output requires. The prefill runs every layer for every prompt token, and every decode step loads and stores whole recurrent states, which on Qwen3.8-Flash-Next take as many bytes per request as the attention cache of thousands of tokens. We find two properties of trained hybrids that locate this excess. First, writes are easier than reads: writing a deep layer's memory of a token needs little of the deep computation that predicting the next token needs. Trained copies of the deep layers' write paths, which we call emitters, therefore write a prompt's deep memories from the residual stream of a shallower anchor layer, and all prompt tokens but the last run as few as 63% of the layers. Second, recurrent states have a state sink: most of their energy lies in a rank-one term along a fixed direction of each head, and the rest is a low-rank working memory. SEED builds on both: it makes the prefill shallow in depth and the decode state shallow in representation. Each recurrent state is stored as its sink plus low-rank factors, which the Gated DeltaNet, Kimi Delta Attention and Mamba-2 updates extend exactly; we prove that the open-loop error of the periodic compressions back to low rank is bounded independently of their number. SEED keeps the original weights frozen, compresses no attention cache, and runs all layers for every generated token. On six checkpoints of five production hybrids of 30B to over 300B parameters, it is nearly lossless on eleven official exams. Served with SGLang on NVIDIA GB300 GPUs on an agentic-coding trace, it raises the per-GPU throughput of Qwen3.8-Flash-Next 1.2-1.4x.

The Harness Is the Model: Agent Scaffolding Changes a Fixed LLM's Compliance with Banking Conduct Rules, Unevenly Across Models

Mehmet Murat Huyut — Thu, 05 Nov 2026 00:00:00 +0000

92% · Evaluation 92%, Agents 87%

Banks validate language models, but they deploy systems: a model inside a harness that decides what it sees, which tools it can call and who checks its actions. We test whether that harness changes compliance with banking conduct rules when the model is held fixed. Four open-weight models each ran five harnesses (a pipeline that makes one structured decision, a ReAct-style tool loop, a loop in which the policy is reachable only through search, a loop with a long benign customer memory, and a loop with a same-model compliance reviewer) on 156 pre-registered cases in a simulated retail bank whose 11 rule families paraphrase EU consumer-credit, payments, investor-protection, data-protection and anti-money-laundering provisions. Every one of the 3120 two-turn episodes was scored deterministically from the executed tool trace, with violations and over-refusals counted separately. Pooled across models the harness moved the correct-case rate only between 87.3% and 92.1%, but within a model it mattered a great deal: for one model the spread between its best and worst harness was 11.5 percentage points, enough to move it from first to last place among the four models, while for another the harness made no detectable difference; the model x harness interaction itself was suggestive rather than conclusive. The same-model reviewer cut violations from 9.6% to 6.4% without raising over-refusals, contrary to our pre-registered prediction. Rules that require date arithmetic failed in every harness. In an exploratory analysis with two LLM judges and a blind human check, 26% of violations were accompanied by a customer message that denied the action the agent had just executed, which a transcript-only review would record as a correct refusal. For model-risk management the unit of validation should be the model-harness pair, and conduct testing should read the action log, not the conversation.

CRESiST: Self-Enhancing Explorative Policy Learning for Simultaneous Speech Translation

Chenyang Le, Pei Zhang, Yuxiang Chen, Bing Han, Derek F. Wong, Baosong Yang — Thu, 05 Nov 2026 00:00:00 +0000

92% · New model 92%, Efficiency 87%, Evaluation 45%

Simultaneous Speech Translation (SimulST) requires balancing translation quality and latency. While Large Language Models (LLMs) excel in offline translation, current LLM-based SimulST systems predominantly rely on prefix-incremental generation. This paradigm suffers from redundant Key-Value cache recomputation and necessitates external Voice Activity Detection to process unbounded streams. To overcome these bottlenecks, we propose an end-to-end interleaved read-write framework built upon an omni-modal LLM. By interleaving speech chunks and text tokens, our architecture achieves full KV cache reuse, natively supporting segmentation-free, infinite-length inference. Furthermore, to learn optimal streaming policies without rigid offline alignments, we introduce a novel Actor-Critic self-enhancement pipeline. An RNN-T Actor dynamically samples read-write trajectories, while an LLM-based Critic evaluates these paths via intrinsic semantic and latency rewards, distilling the optimal policy back to the Actor. Extensive experiments on sentence-level (FLEURS, CoVoST2) and document-level (MuST-C, ACL 60/60) benchmarks demonstrate that our model establishes new state-of-the-art quality-latency Pareto frontiers, substantially outperforming existing cascaded and static-alignment baselines.

Using Experience Twice: Dynamic Process Rubrics for Privileged Token-Level Credit Assignment

Shumao Wu, Kongcheng Zhang, Jianqiang Huang, Baisheng Lai — Thu, 05 Nov 2026 00:00:00 +0000

92% · New model 92%, Agents 86%, Evaluation 39%

Agent trajectories often contain both successful decisions and failed attempts, yet optimization with sparse trajectory-level rewards struggles to identify which critical decisions should be reinforced or suppressed. We observe that meaningful distinctions among on-policy trajectories can be distilled into natural-language process rubrics and used as process supervision signals. Based on this insight, we propose Online Rubric Self-Distillation (ORSD), which leverages self-derived process rubrics for dense token-level supervision, thereby augmenting outcome-centric reinforcement learning. Specifically, the current policy independently generates multiple rubric bundles at each step according to the same group of trajectories, extracting candidate process rubrics from cross-trajectory behavioral differences. A frozen verifier then evaluates all rubric items against each trajectory in the group, and selects one candidate bundle based on bundle-level alignment between rubric satisfaction and task success. ORSD treats the selected rubric bundle as privileged information during training, then compares the log-likelihoods of the original tokens with and without the rubric. The resulting likelihood gap provides token-specific process credit, which is added to the original trajectory-level GRPO advantage through additive advantage shaping. ORSD thus bridges trajectory-level outcome supervision and fine-grained process supervision, enabling token-level credit assignment using process signals. Extensive experiments on ALFWorld and WebShop show that ORSD accelerates policy improvement and boosts final task performance for both Qwen3-1.7B and Qwen3-4B under the same environment-interaction budget. Our code is available at https://github.com/AnnaXChun/ORSD.

Measuring Risk, Not Harm: Detecting Psychological Harm Risk Factors in LLM Conversations

Thu, 05 Nov 2026 00:00:00 +0000

92% · Evaluation 92%, Efficiency 60%

Psychological harms from interactions with large language models cannot be monitored directly because they may emerge over repeated interactions and depend strongly on subjective individual user context. We propose measuring risk, not harm: detecting observable AI behaviors that constitute risk factors for psychological harm at the AI response level and aggregating these signals to support longitudinal monitoring. We operationalize eight such behaviors and construct a synthetic multi-label corpus of 47,062 AI responses and 373,873 response-factor labels, with targeted human annotations for validation and evaluation. We then compare trained embedding-based classifiers, direct LLM judging, and a classifier--judge cascade. All detectors are model-agnostic and operate only on LLM outputs, requiring no access to internal model states. We apply these detectors to several out-of-distribution human-AI conversation data sets from prior research. On held-out human-labeled data, the best cascade achieves macro precision of 0.86, recall of 0.76, and F1 of 0.79. On WildChat, it reduces LLM-judge calls by 97.1\% relative to direct judging. On DelusionEval, risk-factor detections cluster within conversations. Together, our results show that response-level behavioral risk-factor detection can provide a scalable measurement layer for longitudinal psychological-safety monitoring without conflating observable AI behavior with latent user harm.

Objective-Conditioned Active Preference Learning for Multi-Objective Alignment

Duc Anh Vu, Thong Nguyen, Hoang Pham, Ponhvoan Srey, Luu Anh Tuan — Thu, 05 Nov 2026 00:00:00 +0000

91% · New model 91%, Efficiency 64%, Evaluation 42%

Aligning large language models (LLMs) with several objectives, such as helpfulness, harmlessness, and honesty, typically relies on preference judgment for each objective, growing the annotation cost linearly with the number of objectives. Active preference learning can reduce this cost by selecting the comparisons whose labels are expected to change the policy most. Existing remedies either lack an objective-aware query selection mechanism for the multi-objective setting or regard conflicting preferences as harmful, therefore excluding them from the training data. We introduce OGC-DPO (Objective-conditioned Gradient Coverage DPO), an active learning framework that treats each (response pair, objective) combination as a separate query and measures its informativeness through the policy's own implicit gradient, without relying on an external reward model. OGC-DPO then selects the queries that most expand each objective's gradient coverage. Empirical experiments across varying numbers of objectives demonstrate the effectiveness of OGC-DPO over existing active preference learning methods under the same annotation cost. Further analysis demonstrates that informative conflicts, rather than being uniformly harmful, provide a useful policy training signal.

SafeStart: Mitigating the Overlooked Safety Cost of Active Expert Reduction in MoE LLM Serving

Thu, 05 Nov 2026 00:00:00 +0000

91% · Efficiency 91%, Evaluation 61%

Mixture-of-Experts (MoE) language models send each token through a few of many expert subnetworks. Serving methods make these models faster by running even fewer experts per token, and they are validated on accuracy benchmarks. We show that these methods can weaken the model's refusal of harmful requests, and we introduce SafeStart, which largely restores refusal while keeping most of the speedup. In vLLM on one RTX A6000, keeping about two thirds of the selected experts speeds up offline decoding by 15 to 28% at batch size 8. We find that, at this setting, under a jailbreak that tells the model not to refuse, the share of harmful requests that the model answers rises by 18.4 percentage points on Qwen1.5-MoE, 9.5 on OLMoE, and 6.6 on DeepSeek-V2-Lite, the three of nine models where it rises significantly, while MMLU accuracy drops by less than two points. We discover that most of this increase arises while the model processes the prompt and generates its first few tokens. SafeStart therefore runs all experts for the prompt and the first eight generated tokens, and the reduced set of experts afterwards, through a fused kernel in vLLM. With experts removed, SafeStart keeps at least 85% of the offline decode speedup at batch sizes 8 and 32 and brings the change under this jailbreak to between -2.1 and +1.3 points for three expert-reduction methods on these three models. On 889 held-out requests, in an analysis fixed before generation, truncation raises harmful compliance by 5.8 to 21.4 points, and SafeStart leaves a rise of 0.6 to 2.6 points, non-inferior to the original model on two of the three models. Our results show that faster MoE serving need not cost safety: a serving system can keep most of the gain from running fewer experts by running all of them at the start of each response, and it should test refusal, not only accuracy, before it enables such a method.

GIST: Gibbs-Targeted Hierarchical Data Selection for Offline RLVR

Xuliang Huang, Tianxiang Zhao, Chengwei Qin, Yao Shu — Thu, 05 Nov 2026 00:00:00 +0000

91% · New model 91%, Efficiency 61%, Evaluation 41%

Offline reinforcement learning with verifiable rewards (RLVR) trains on a fixed pool of sampled responses and typically supervises every token of every response, although responses and tokens are not equally informative. We study which ones to supervise through the Gibbs-optimal policy of KL-regularized RLVR, which offline training targets, and measure the mismatch between a policy and this target with a bidirectional KL divergence. At the reference policy, this mismatch factorizes exactly and separates the two outcomes: correct responses contribute mainly through the forward term, concentrated on their low-probability tokens, and incorrect responses through the reverse term, concentrated on the responses the policy generates often and on their high-probability tokens. Based on this analysis, we propose GIST (Gibbs-Targeted Hierarchical Data Selection), a two-level selection method that approximates these principles with signals available before training. GIST uses the sampling temperature to select correct responses on-policy and incorrect responses near the policy's mode, and it uses the initial policy's token probabilities to supervise only the tokens where the mismatch lies. Applied to two offline training objectives, GIST raises the average accuracy over GSM8K, MATH-500, and MBPP for all three models under both objectives. In a controlled comparison on Qwen3-0.6B/GSM8K, it outperforms seven selection baselines that choose from the same candidates while supervising 66.6% fewer tokens than training on all candidates of the same prompts.

EvoOPD: Evolving Privileged Information Strategies for Black-Box On-Policy Distillation

Ximo Zhu, Ruiqi Liu, Yi Han, Zhengan Wu, Gengsheng Li, ZiAn Wang — Thu, 05 Nov 2026 00:00:00 +0000

91% · New model 91%, Evaluation 35%

Proprietary language models exhibit strong reasoning and coding capabilities, but their output distributions are inaccessible, making fine-grained distillation methods that rely on teacher logits difficult to apply directly to black-box models. We observe that rich textual feedback from a black-box teacher can be treated as privileged information (PI) and converted, through a PI-conditioned privileged teacher, into dense supervision signals on the student's own trajectories. However, what PI should be acquired from the teacher to most effectively support student learning remains unclear. PI strategies that differ in where they focus, what information they request, and how much detail they provide can have substantially different effects on the current student, and their relative effectiveness can further change as training progresses. Based on this observation, we propose Evolving Privileged Information Strategies for Black-Box On-Policy Distillation (EvoOPD), which treats PI acquisition itself as an optimization problem and maintains a PI Strategy Bank that evolves with the student state. EvoOPD uses the measured repair utility of PI on the current student to adapt the focus, information type, and granularity of teacher feedback, allowing the acquired information to continually match the student's learning needs. We systematically evaluate EvoOPD on mathematical reasoning and code generation with Qwen3-1.7B, 4B, and 8B students. EvoOPD achieves the best overall performance in all six model-task settings, improving over the strongest baseline in each setting by an average of 2.14 percentage points in mathematics and 1.30 points in code. Further analysis shows that the relative effectiveness of PI acquisition strategies changes with student state, making a fixed strategy difficult to sustain as the best choice throughout distillation.

LooM: Loop-Invariant Memory for Causal Encoder-Decoder Language Models

Lele Yang, Muxi Diao, Zhonghao Yan, Jiankang Chen, Bin Wen, Tingting Gao — Thu, 05 Nov 2026 00:00:00 +0000

91% · New model 91%, Efficiency 48%

Looped Transformers increase effective depth by re-executing shared layers, but three limitations hinder their adoption: memory grows with the loop count, compute can only be adjusted in whole-block steps, and the loop count is hard to change after training. We propose LooM (Loop-invariant Memory), which addresses each in turn. In a causal encoder-decoder (CED), LooM loops only the decoder layers after the first, which read a global memory written once per token, and each of them publishes the key-value (KV) entry for a token only on its first visit, so the cache does not grow with the loop count. These layers loop independently, providing every budget from 3 to 21 layer executions per token, and one model trained on randomly sampled budgets covers all of them without a router. With enough loops, LooM outperforms a single-pass baseline trained on the same data on language modeling, zero-shot accuracy and long-context perplexity, while using the same cache size. Our analysis further shows that the training budget distribution decides what extra depth buys. These results suggest that a CED is a natural place to loop: memory is written once, and depth becomes a budget chosen at deployment.

Verified but Vacuous: Three Failure Surfaces of LLM-Driven Verified Code Generation

Thu, 05 Nov 2026 00:00:00 +0000

90% · Evaluation 90%

What should ``verified'' mean when evaluating LLM-generated code? Many existing pipelines report a single verification number. We argue this collapses three distinct properties measured in a conditional audit sequence: verifier acceptance (Surface A), \emph{syntactic specification binding (Surface B: does the proof obligation bind to the spec?), and \emph{semantic specification fidelity} (Surface C: does the spec capture the informal requirement?). Two empirical demonstrations show the surfaces separate. \textbf{First ($A \neq B$), via Li@Opus (a same-model reproduction of \citet{li2025dafny}'s Dafny-as-intermediate-representation (IR) pipeline under Claude Opus 4.7) on three Python benchmarks}: it reaches \textbf{99.1\% verifier acceptance} on HumanEval while only \textbf{1.9\%} carry a certificate binding to a spec-defined predicate or function. \textbf{Second ($B \neq C$), via VCG on two Rust benchmarks}: we use VCG (which is our own in-house Dafny-as-intermediate pipeline) on Rust because the Rust target lets us enforce fully-verifiable output (every component ships as a Dafny certificate, ensures-binding gate passing by construction). Even inside that construction, \textbf{11.48\% (MBPP-Rust) and 14.82\% (VeriContest LC slice) of structurally-verified samples still fail hidden tests}. On the LC slice of VeriContest, \textbf{VCG's reported End2End rate is 7.79\% (535/6870).} The cited external rates are 5.51\% (GPT-5.5) and 1.88\% (single-shot Opus 4.7). These give descriptive differences of 2.3 percentage points and 4.14$\times$, respectively; task-budget, repair and harness differences prevent a matched-performance claim.

Reward Budgeting Reduces Premature Convergence in Reinforcement Learning for LLM Reasoning

Mengni Jia, Mengyu Zhou, Xiaoxi Jiang, Guanjun Jiang — Thu, 05 Nov 2026 00:00:00 +0000

90% · New model 90%

RL for LLM reasoning often reaches an early performance plateau as the policy distribution becomes increasingly sharp. Recent work has studied policy entropy as both a diagnostic and a control signal for this behavior. In our experiments, however, increasing entropy alone does not reliably change the eventual plateau. We analyze RL through its implicit target policy and derive a sharpness factor $\kappa$, together with an effective counterpart $\kappa_{\mathrm{eff}}$, that describes how strongly reward or advantage gaps are translated into probability-ratio gaps. Motivated by this view, we introduce \emph{reward budgeting}, a plug-in advantage construction that assigns a fixed total reward budget to each prompt. For binary rewards and independent on-policy rollouts, its unclipped score-function estimator follows the gradient of an exact finite-$G$ concave objective with diminishing returns, while its advantage gap decreases as more rollouts become correct. For general rewards, a shifted log-mean interpretation holds asymptotically under the stated regularity conditions. These results motivate, rather than establish exact objective equivalence for, the clipped implementations. On Qwen3-8B and Llama3.2-3B-Instruct, reward budgeting consistently improves a range of RL baselines, with gains up to +12.50 on AMC23 and +10.00 on AIME24 and AIME25. Overall, our results support an update-dynamics view of premature convergence and show that strong gains can be obtained without explicitly preserving entropy.

Familiar letters, not related languages, let unseen languages into a shared space of meaning

Yuliang Liu, Xinyu Wang, Xiang Bai — Thu, 05 Nov 2026 00:00:00 +0000

90% · New model 90%, Evaluation 34%

Hundreds of languages have written text but almost no translations, and whether a model learning one from untranslated text builds on a related language or familiar letters is unclear: they usually coincide. Here we show that familiar letters, not a related language, let an unseen language into a shared space of meaning. GlyphMind, a tokenizer-free vision transformer that reads text as pixels and lets any sentence be shown in any script, was pretrained to align translations across 27 writing systems. Adapted on untranslated Lao, a language and script it had never seen, it retrieved English translations seven times as often as the strongest simple baseline. Rewritten in unfamiliar letters, Lao did not enter although its close relative Thai was in pretraining; Santali, which had none, entered in Devanagari, familiar from Hindi, but not in its own script. Transliteration into familiar scripts may thus open multilingual models to languages without translations.

Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription

Daniel Huang, Qi He, Shaoxiong Duan, Yingjie Li, Qiao Cheng, Yitao Duan — Thu, 05 Nov 2026 00:00:00 +0000

90% · New model 90%, Evaluation 41%

Streaming speech recognition is increasingly consumed by machines, such as translators and voice agents, that act on each token as it appears. A token that is later revised costs wasted work or a wrong action. In this work, we present Confucius4-R2T2 (short for Real Real-Time Transcription), a stable-prefix streaming ASR model that never modifies its earlier output while maintaining high accuracy and low latency. R2T2 is obtained by post-training Qwen3-ASR-1.7B on 1,500 hours of Chinese and English audio while freezing its pre-trained audio encoder and projector. The adaptation is based on the concept of Longest Stable Prefix (LSP), a new notion that we introduce in this work. We prove that LSP is monotone under the prefix order, consistent with the full-audio output, and maximal among labels that equal a whole earlier output. Learning from LSP labels converts a full-context model into a stable-prefix streaming one without any architectural change. Extensive evaluations show that R2T2 brings streaming recognition close to the base model's full-context quality while delivering stable text with low emission lag. Its gains extend beyond the post-training languages, demonstrating a practical path from full-context ASR to stable streaming. We open-source the model weights and code at https://huggingface.co/netease-youdao/Confucius4-R2T2 and https://github.com/netease-youdao/Confucius4-R2T2.

Learning When to Think Aloud in Latent Reasoning

Zhehao Zhou, Tongrui Su, Shengyu Zhu, Da Li, Zeyu Liu, Jun Zhu — Thu, 05 Nov 2026 00:00:00 +0000

90% · New model 90%, Efficiency 76%

Latent Chain-of-Thought (CoT) reasoning improves efficiency by shifting intermediate computation into continuous hidden states. However, fully latent methods often struggle on precision-critical tasks such as mathematical reasoning, while existing hybrid approaches retain substantial token-generation overhead. We introduce VITAL, an importance-guided hybrid reasoning framework that selectively verbalizes informative tokens as reasoning anchors while keeping most computation latent. A lightweight gating head dynamically selects between explicit token generation and direct hidden-state feedback. VITAL constructs hybrid trajectories from explicit CoT using token- importance estimation and trains the model through curriculum-based supervised training, followed by reinforcement learning with a reward that balances correctness and conciseness. Experiments show that VITAL consistently out- performs existing fully latent reasoning methods, achieving relative average accuracy gains of 9.6% on GPT-2 and 5.1% on Llama-3.2-1B at comparable reasoning lengths. Compared with hybrid baselines, VITAL achieves higher reasoning accuracy with substantially shorter reasoning trajectories on Qwen3-1.7B.

LegalEvolver: Agentic Legal Judgment Prediction via Rubric-Guided Self-Evolution

Xinyu Yang, Jiejun Tan, Chenlong Deng, Chenghao Zhang, Tongyu Wen, Zhicheng Dou — Thu, 05 Nov 2026 00:00:00 +0000

90% · Agents 90%, New model 87%, Evaluation 44%

As LLM-based agents develop, they have been increasingly applied in the legal domain. Legal Judgment Prediction~(LJP) requires models to conduct multi-step reasoning and external legal evidence retrieval. Existing legal search agents' performance is often constrained by the quality of retrieval queries. They often formulate queries directly from the initial user request, overlooking latent legal issues and legally salient details. To address this issue, we propose \emph{LegalEvolver}, a query optimization framework specifically designed to generate legally informative queries for LJP-oriented retrieval. Sampling high-quality optimized queries and their corresponding search trajectories is challenging, as relying solely on unguided high-temperature rollouts is both inefficient and constrained by the model's initial capabilities. We therefore heuristically optimize queries as seeds, improving the efficiency of trajectory generation. Furthermore, we summarize rollout experience into rubrics and use them to guide multi-round iterative self-evolution. Iteratively performing rubric optimization and rubric-guided RL enables the model to surpass its initial performance ceiling. End-to-end performance on extensive benchmarks demonstrates the effectiveness of query optimization on LJP. Comparisons with direct SFT followed by RL demonstrate the effectiveness of rubric guidance. Evolving rubrics further outperform fixed rubrics, confirming the benefits of self-evolution.

★ SLATE: A Large-Scale Benchmark for Long-Horizon API Selection in E-commerce Tool Use

Rongzhe Wei, Ge Shi, Min Cheng, Sarthak Ghosh, Vaibhav Gorde, Leman Akoglu — Wed, 04 Nov 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 97%

Enterprises increasingly want LLM agents to automate standard operating procedures over their software stacks -- mapping each step of a procedure to the right call among thousands of internal APIs, without hand-authored step-to-API mappings. The core difficulty is API selection at scale: choosing and composing the correct tools over a long horizon, judged only by whether the overall task succeeds. Current tool-use benchmarks do not test this regime -- they expose only a handful of tools, evaluate single-step invocation, or grade multi-step behavior with subjective LLM-as-a-judge scores that penalize valid alternative trajectories. We introduce SLATE (Synthetic Large-scale API Toolkit for E-commerce), a large-scale benchmark that tests large-toolset, long-horizon API selection in a realistic enterprise domain. SLATE couples (i) queries decomposed into hierarchical, tool-grounded procedures over a ~1,000-tool e-commerce library, (ii) a deterministic, context-aware simulator that returns recorded outputs for matching calls and default responses otherwise, and (iii) an end-to-end task-success metric -- a task succeeds when it reaches every reference outcome -- that credits diverse-but-valid trajectories while remaining fully automatic and reproducible. Experiments with six open decision models and the language models they are built on, from 3B to 12B parameters, show that even the best model completes only 5.0% of the tasks, establishing SLATE as a challenging testbed for reliable enterprise tool use. We release SLATE, its simulator, and the evaluation harness at https://geshijoker.github.io/slate/

★ Do Multi-Turn Conversational Capabilities Carry Over from Text to Speech? A Parallel Multilingual Benchmark

Jiaming Zhou, Pei Zhang, Shiwan Zhao, Zhijun Wang, Baosong Yang, Yong Qin — Wed, 04 Nov 2026 00:00:00 +0000

99% · Evaluation 99%

Successful multi-turn voice interaction requires models to retain information from earlier turns, follow persistent instructions, and track revisions to conversational state. These abilities are usually evaluated from transcripts, leaving open whether successes demonstrated from text carry over when the same turns are spoken, and whether this transfer is consistent across languages. We introduce Omnilingua-AudioMC, a strictly parallel multilingual extension of Audio MultiChallenge, to study this question. The benchmark contains 354 conversations and 1,404 rubric items across ten languages, with content, turn structure, cross-turn dependencies, and evaluation criteria aligned across transcript and speech conditions. Across six speech-language models, the shift from reference transcripts to direct speech input is strongly model- and language-dependent: models with similar transcript performance can diverge substantially under speech, and their relative strengths across languages can change. Paired rubric-level analysis shows that aggregate transcript-speech gaps can obscure which conversational requirements are preserved, because losses of transcript-demonstrated successes can be offset by gains on other items. Replacing preceding user turns, the current turn, or both with model-generated transcripts localizes recoverable degradation: for the two models with the largest gaps, history substitution yields a larger macro-level improvement than current-turn substitution in all ten languages, while the strongest direct-speech model gains little from either. These results show that transcript performance is an unreliable proxy for multilingual spoken conversational ability and identify the representation of preceding spoken context as an important source of recoverable degradation. Our code is publicly available at https://github.com/QwenLM/Omnilingua-Bench/tree/main/Omnilingua-AudioMC.

★ Are LLM Assistants Regulatory Ready? Answers from regenold's 2026 EU AI Act Q&A Challenge

Marco Virgolin, Thomas Haszprunar, Maximilian Regenold — Wed, 04 Nov 2026 00:00:00 +0000

99% · Evaluation 99%, Hallucination 61%

Large language model (LLM) assistants are increasingly proposed for regulatory affairs, yet their reliability in regulatory tasks remains unclear. We introduce the EU AI Act Q&A Benchmark, an expert-curated benchmark of 110 open-ended questions about the EU AI Act (Regulation (EU) 2024/1689). Each question comes with an exemplary answer, question-specific correctness criteria, and expected references. Evaluation measures answer correctness, reference quality, conciseness, tone, and latency, with a multi-turn ``hard mode'' testing responses to simulated user pushback. We also present the results of the 2026 EU AI Act Q&A Challenge, based on the benchmark, where three off-the-shelf baselines and five purpose-built assistants (from four commercial developers and one academic entrant) were tested. Only two of the five submitted assistants consistently outperformed the baselines; one achieved high scores across all evaluated metrics. These results show that specialist development alone does not guarantee better performance and highlight the critical importance of LLM assistant evaluation in regulatory affairs. The benchmark code is openly available at https://github.com/regenold-GmbH/eu-ai-act-benchmark-2026/.

★ IdeaTaste: Do LLMs Share the Research Taste of Human Researchers?

Wed, 04 Nov 2026 00:00:00 +0000

99% · Evaluation 99%

Research taste, the judgment of what to read, which problems to pursue, and which methods to try, increasingly shapes the work of AI scientists, yet it remains far less understood than their ability to execute research. We introduce IdeaTaste, a benchmark that scores an LLM by how closely the research proposals it writes from a literature snapshot frozen at 2025 match the work the same communities published in 2026. Our experiments span six contemporary language models and four research areas from established to emerging fields, covering more than 90{,}000 proposals in total. LLM agents fall well short of human communities in both the literature they cite and the ideas they state, reaching about 26\% of the human level on citation matching and only 18\% on idea matching. Their populations can cite as diversely as a community and take up its tasks and methods, yet rarely arrive at the concrete approaches and research plans it pursues. Neither richer field context nor larger populations substantially close this gap. In blinded expert review, AI proposals trail the 2026 papers least in presentation and most in expected success, echoing their gap in concrete approaches, and those whose ideas no human paper shares fare worst, suggesting weaker directions rather than overlooked opportunities. These results indicate that the main bottleneck in AI research taste lies less in finding the directions a community works on than in turning them into the concrete approaches it goes on to pursue. We release IdeaTaste as an open, continuously updated leaderboard (https://cxcscmu.github.io/IdeaTaste) that benchmarks contemporary AI proposals against future publication cycles.

★ Unsatz: Evaluating Scientific Discovery Without a Given Ansatz

Ethan Ji, Wei Wang, Yizhou Sun — Wed, 04 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 31%

Before a scientist fits any parameter, they must decide what the model contains: which quantities exist, which are hidden, how they interact, and how they are measured. Physicists call such a trial form an ansatz. Benchmarks for automated equation discovery mostly supply it, by fixing the variables, a library of candidate terms, or a law to be modified. We introduce Unsatz, a benchmark in which an LLM agent must construct the ansatz itself. Each task is a simulated system, generated from a private seed as a program in a small modeling language: ordinary, delay, or partial differential equations, discrete maps, static and causal models, or stochastic processes, possibly with hidden states, distorted sensors, and uninformative measurement channels. The agent buys experiments within a budget, writes candidate programs in the same language, obtains their coefficients from the benchmark's fitter, and is scored only on how its final program predicts experiments it never saw: conditions outside the range it could explore, interventions, and longer runs. Across 80 systems in 20 domains, nine language models solve between 21% and 60%. On 44 of these systems, classical equation-discovery methods almost never succeed, although the same data usually suffice once the true equations are given with unknown coefficients; the difficulty lies in finding the structure, not in estimating it. Because mechanisms recur across systems, we also give agents the program they discovered for a related system; it helped only when the new system shared the mechanism and the first investigation had succeeded. Stronger agents make fewer unscientific errors; beyond these, agents find which variables a system has but not how they act, and prefer revisions that lower the training error to revisions that predict better.

★ EMBERS: Empirically Motivated Robustness Testing of Deep Research Systems via Defect-Guided Fuzzing

Shuochuan Li, Zan Wang, Dong Wang, Yuhang Jiang, Huajin Cui, Chen Zhang — Wed, 04 Nov 2026 00:00:00 +0000

98% · Agents 98%, Evaluation 95%, Hallucination 45%

Deep Research (DR) systems extend large language models into autonomous agents that decompose complex information needs, iteratively search the web, and synthesize evidence to answer challenging questions. However, their multi-stage execution may introduce robustness risks, as variations in user questions or intermediate search inputs may propagate across research stages and lead to incorrect answers. Existing evaluations focus on task-solving capabilities or robustness against external evidence disturbances, leaving intrinsic weaknesses in DR execution insufficiently explored. Thus, we conduct an empirical study of 1,932 execution trajectories from three representative DR systems and identify nine recurring defect categories across subquestion formulation, online search, and final reasoning. Guided by these findings, we propose EMBERS, an empirically motivated, defect-guided fuzzing framework for robustness testing of DR systems. Starting from correctly answered seed questions, EMBERS generates stage-targeted test inputs using defect-guided mutation, web-search-assisted fidelity assurance, and diversity-aware seed maintenance. Experiments across 16 framework--model--dataset settings show that EMBERS achieves 3.21 times the failure-discovery throughput of the strongest baseline under identical testing budgets. Human and LLM-assisted evaluations demonstrate that most generated inputs preserve the original information need and necessary contextual constraints. Moreover, model fine-tuning experiments show consistent-answer performance improvement, highlighting their value for robustness testing and model improvement.

★ Typological Feature Induction as a Benchmark for Inductive Metalinguistic Reasoning

Chen Zhang, Seth Aycock, Yusuke Miyao — Wed, 04 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

Existing metalinguistic reasoning benchmarks require both inference of rules and their application, making failures hard to diagnose. We introduce typological feature induction (TFI) which targets the former step. Given sentences in an unnamed language paired with translations in a known pivot language, models must predict a typological feature of the unnamed language. From Grambank and multi-parallel corpora, we automatically build evaluation questions over 195 languages. We find that LLMs identify 59--80% of these languages from only 16 sentence pairs; and misaligning the translations barely affects their accuracy, suggesting reliance on prior knowledge of the language. Enciphering text to prevent recognition increases the corpus' utility, but mainly for features where target and pivot languages agree. On languages which models cannot identify, neither open model outperforms a majority-value baseline. Since induction from raw text aligns with field linguists' use-case, this work indicates the models tested are not yet ready for this role. TFI offers a scalable path to track progress in this direction.

★ CardioVascBench:Prioritized Diagnosis Generation from Real-World Chinese Cardiovascular Records

Minghao Kong, Muyao Chen, Jianhao Wu, Yuqiang Zhang, Xiangbin Meng, Zirong Liu — Wed, 04 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

Identifying the primary diagnosis alone does not fully characterize a model's diagnosis-generation performance. We introduce CardioVascBench, a retrospective benchmark for prioritized diagnosis generation from real-world Chinese cardiovascular inpatient records. It contains 3,262 de-identified cases, with 2,174 for training and 1,088 for evaluation, linking clinical narratives and textual ECG reports to expert-verified, ordered discharge diagnoses standardized to ICD-10. An English translation of the evaluation set supports aligned cross-lingual comparison. Our evaluation protocol combines terminology normalization and rank-aware scoring to separately assess primary-diagnosis identification, reference-diagnosis coverage, and list precision. Evaluation across 36 systems shows that higher primary-diagnosis accuracy does not necessarily imply higher list precision against discharge references. We also provide CardioKG-Agent, a knowledge-enhanced baseline combining entity normalization, constrained graph retrieval, and guideline evidence. In a staged comparison of three baseline models, supervised fine-tuning plus the agent achieves a 73.5% higher mean composite score than fine-tuning alone, representing a system-level gain. These results measure retrospective agreement with recorded diagnoses; agreement between metric rankings and physician judgments remains to be validated. CardioVascBench supports research on diagnosis prioritization and knowledge-enhanced generation.

★ Kierston & KierstonBench: Simulation of Allentown, PA and Evaluation of Language Models in Crime Investigation and Surveillance Tasks

Andrew Semenov, Svyatoslav Dorofeev — Wed, 04 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 70%

Large language models are candidates for collecting and analyzing records in criminal investigations and mass surveillance. Their capability and accuracy in this work cannot currently be measured. Real deployments lack the ground truth needed to score a model's conclusions, and no benchmark evaluates these capabilities in a simulated environment. We present Kierston, a city-scale simulation engine, and KierstonBench, a benchmark that tests these capabilities against known simulated events. Kierston models approximately 125,000 residents of Allentown, Pennsylvania, generating their daily activities and the telecommunications, financial, administrative and license-plate records that surveillance and investigative systems collect about them. These records derive from a common behavioral timeline and are distributed across sources representing separate providers. Models investigate through a restricted records interface, linking observations across sources without access to the underlying simulation. The benchmark comprises 32 task families covering record linkage, routine reconstruction, incident investigation and assessment of deceptive accounts. In an evaluation of sixteen language models, most could not perform records investigation and submitted identifications the records did not support. A small number of frontier models solved most of the tasks and abstained when evidence was insufficient. In some cases these models still fell for common deceptions and assumptions, and none replaced an investigator.

★ AgenticFinBench: Benchmarking Financial Agents on Diverse Tasks from Real-World Requests

Shu Zhou, Ziyang He, Yuxin Rao, Yudong Fang, Jiali Qian, Dongyuan Chen — Wed, 04 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 97%, Agents 80%

Financial agents are becoming an important class of real-world AI systems. Financial tasks vary widely in type and difficulty, and many are open-ended with no ground-truth answers. However, existing benchmarks neither capture this breadth and open-endedness nor separate grounding from correctness. To bridge this gap, we introduce $\textbf{AgenticFinBench}$, a benchmark of 230 real-world financial tasks spanning 8 categories, 2 languages, and 4 complexity levels. We also propose a reference-free evaluation framework that combines question-specific rubrics with a two-stage claim audit that detects hallucinations, defined as claims absent from the agent's trace and refuted by external evidence. This framework better matches how financial work is assessed in practice. Experiments across 18 agent configurations reveal three key findings: (1) Ignoring hallucinations can overstate agent performance: hallucination auditing lowers the baseline leader's score by 7.58 points and moves it to fourth place. (2) Harness differences explain more variation in evidence grounding than model differences (28.6% vs 8.8%). (3) Quality scores fall from 89.44 on direct retrieval to 79.18 on the most complex tasks. Case studies show that agents struggle to explain how events affect company profits or stock prices, compare financial data on a consistent basis, and justify their conclusions. By releasing AgenticFinBench, we aim to advance financial reasoning, factual verification, and agent design for more reliable responses to diverse real-world financial requests. Our project page is at $\textbf{https://nashnovaresearch.github.io/AgenticFinBench/}$.

★ One Assistant, Many Memories: Benchmarking Real-World Group Assistance When Recall Is Not Enough

Zhanghao Hu, Linhai Zhang, Qian Zhao, Qi Zhu, Yuan Hua, Di Liang — Wed, 04 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 34%

The value of memory lies not only in recall but in using it to provide appropriate assistance. Recent benchmarks test memory use in personalisation and agentic tasks, but seldom evaluate assistance without fixed answers or integration across multiple participants. Group chats make both challenges more acute: an assistant must integrate information from multiple members, track revisions and conflicts over time, and produce one response for the group. Yet existing group memory benchmarks rely on synthetic conversations and inherit constrained evaluation from general memory benchmarks, focusing on recall against fixed answers or decisions over a closed set. We introduce GroupAssistBench, the first memory benchmark from real-world group conversations with a deployed group assistant, covering 80,299 messages in 50 groups and costing over $53K. Over the same histories, GroupAssistBench evaluates memory recall (L1) with reference answers and memory use (L2) with service requests without fixed answers across single, distributed, and conflicting memories (M1--M3). Requests are verified to require group history, memories are anchored to their sources, and L2 uses tailored rubrics with penalties derived from real user feedback. Across evaluations of frontier models with full context, retrieval methods, and memory systems, the strongest memory system rivals full context on recall but falls far behind on use, especially for distributed and conflicting memories. Together, two additions that relate retrieved memories and bind remembered facts as response constraints bring consistent gains, showing that group memory systems must improve use, not recall alone.

★ ScamWorld: An Interactive Environment for Evaluating Web Agents Against Online Fraud

Yutao Wu, Xiao Liu, Ming Wen, Hanxun Huang, Jun Wang, Xingjun Ma — Wed, 04 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 97%

Web agents increasingly browse, shop, and manage accounts on behalf of their users. Prior work has shown that they can be misled by adversarial content, including prompt injections and deceptive interfaces. Beyond such attacks, agents operating on the open web also encounter online fraud designed to deceive humans, including fraudulent online stores, fake investment platforms, and advance-fee schemes. How web agents respond to such fraud during task execution remains underexplored, in part because scam websites are short-lived, making systematic evaluation difficult. To address this gap, we introduce ScamWorld, an interactive environment that reconstructs reported scam websites across diverse scenarios, fraud mechanisms, and languages. Each reconstructed website reproduces documented fraud mechanisms, making relevant signs of fraud observable before the agent's final action and enabling controlled, repeatable evaluation. Through an extensive study of web agents powered by six frontier LLMs, we find that baseline deception rates range from 11% to 35% across models. Agents proceed more readily with payments than with requests for credentials or wallet information. A closer analysis of agent trajectories shows that some agents acknowledge suspicious conditions yet proceed, at times treating the user's instructions as authorisation to accept additional terms imposed by the website. ScamWorld provides a reusable testbed for systematically evaluating and stress-testing web agents against real-world online fraud.

★ ToolSpace: Scaling Tool-Centric Environment Synthesis for Long-Horizon Agents

Guanting Dong, Zhichao Sun, Xinyu Yang, Rui Wang, Xin Deng, Jiajie Jin — Wed, 04 Nov 2026 00:00:00 +0000

98% · Agents 98%, Evaluation 93%

Large language models are deployed as long-horizon agents that compose diverse tools across MCP services into executable workflows. Existing MCP environment-scaling methods are data-centric: they first synthesize a database and scenario, then derive tools and tasks from them. As a result, tool interfaces may not match real ones, tool compositions are not grounded in execution, and tasks rarely span more than one environment. We propose ToolSpace, a tool-centric framework that reverses this order. ToolSpace mines data, state, and retry dependencies from adjacent calls in execution traces of real MCP services and open agent corpora, using deterministic rules rather than LLM judgment, and clusters the resulting weighted tool-composition graph into environments. Databases are synthesized only for state-mutating environments, while query and stateless tools replay recorded returns. ToolSpace then performs state-aware random walks with rollback and decompiles each chain into a natural-language task with assertion-based checks, keeping only samples that replay successfully from the initial state. Applied to AutomationBench and tau3-Bench Banking, the pipeline produces harder Pro versions on which 7 frontier models drop by up to 38.2 points. Across seven long-horizon tool-use benchmarks, ToolSpace raises Qwen3.6-35B-A3B from 23.0 to 41.0 average score (43.2 with RL) and GLM-5.2 from 44.6 to 56.6, on par with Claude Opus 5. Ablations show that execution-witnessed edges clearly outperform LLM-judged ones, and performance keeps improving with data and environment scale.

★ GM League Arena: An Interactive Long-Horizon Task Benchmark Based on Professional Sports League Management

Junkai Zhou, Ziming Pan, Zhaoyi Zhang — Wed, 04 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

Long-horizon decision-making under sustained competition requires large language models (LLMs) to balance immediate investment with future resource needs. However, local success in multi-step interactions can not adequately capture cumulative decision effects or demonstrate long-term goal achievement under resource constraints and sustained competition. To address this gap, we introduce GM Bench, a benchmark for managing professional sports teams across seasons under resource constraints and competition among teams. We build the environment simulators for four sports and add financial constraints, player demands, and market mechanisms to establish a unified task across sports. Diverse rosters of comparable strength are constructed from real-world player data, and each LLM is evaluated as a team general manager to assess its long-term management capabilities. We evaluate competitive performance, cash reserves, and cross-season improvement across four sports. Over five NBA seasons, Claude Opus 5.5 ranks first among 29 LLMs, while GPT-6 Astra and GPT-6.1 Sol combine strong performance with larger cash reserves.

★ JHiMA-WebRL: A Joint Hierarchical Multi-Agent Reinforcement Learning Framework for Long-Horizon Web Tasks

Shanka Subhra Mondal, Jiayun Wang — Wed, 04 Nov 2026 00:00:00 +0000

98% · Agents 98%, New model 86%, Evaluation 33%

Multimodal web agents have shown promising capabilities for interacting with websites through graphical user interfaces (GUIs), yet realistic web tasks often require long-horizon, multi-step planning that remains challenging for monolithic policies. Such agents must jointly infer high-level plans, navigate through intermediate states, track progress over extended interaction histories, and execute visually grounded actions. Although prior work has explored modular multi-agent systems with a separate planner, the planner is typically not updated jointly with the actor. We investigate whether explicitly separating and jointly training planning and execution provides a stronger inductive bias for long-horizon web navigation. We introduce JHiMA-WebRL, a joint hierarchical multi-agent reinforcement learning framework for web agents that couples a task decomposer with an actor. To train the decomposer, we use a strong teacher model to construct structured subtask supervision from multimodal GUI trajectories, and use these annotations to fine-tune both the decomposer and the subtask-conditioned actor. We then extend Group Relative Policy Optimization (GRPO) to jointly optimize the decomposer and actor in the interactive WebArena environment through hierarchical rollout grouping and role-specific advantages. On WebArena-Lite-v2, JHiMA-WebRL-14B, consisting of a 7B actor and a 7B decomposer, achieves an average success rate of 37.2%, compared with 24% for a ScaleCUA-32B agent that we fine-tune with RL on the same training tasks and evaluate under the same protocol. Ablations show that jointly updating the decomposer during RL outperforms updating the actor alone. These results suggest that jointly learned role factorization with smaller models can be a competitive alternative to scaling monolithic agents for long-horizon web tasks.

★ CrossSWE: Benchmarking Cross-Repository Joint Development for Coding Agents

Silin Chen, Yuling Shi, Heng Lian, Zihan Zhang, Xiaodong Gu, Haotian Lin — Wed, 04 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 38%

As coding agents achieve increasingly high success rates on software engineering benchmarks, recent efforts have sought to increase task complexity through longer horizons and more extensive code modifications. However, real-world software development often spans multiple interdependent repositories, introducing additional challenges in dependency reasoning, change propagation, and interface consistency. Rather than further increasing complexity within individual repositories, we introduce CrossSWE, a benchmark for cross-repository joint development comprising 171 real-world instances spanning 165 repositories, 20 technical domains, and 13 programming languages. CrossSWE verifies the necessity of coordinated changes by testing whether partial repository modifications are insufficient to achieve the intended behavior. Our evaluation of five LLMs reveals substantial challenges, with the strongest model, GPT-6.1 Sol, achieving only 58.48% Pass@1. Furthermore, cross-repository inconsistencies emerge as the most prevalent failure mode (44.2%), highlighting the difficulty of maintaining consistency across interdependent repositories.

★ Dory1.0-4B: A Looped Normalized-GPT

Wed, 04 Nov 2026 00:00:00 +0000

97% · New model 97%, Efficiency 67%, Evaluation 30%

We introduce Dory1.0-4B, a 4-billion-parameter looped (latent-recurrent) language model with elastic depth: a single checkpoint serves different inference compute budgets. Dory1.0-4B is built on a normalized Transformer (nGPT) backbone with interleaved sliding-window and full-attention layers, organized into an input block, a shared recurrent block, and an output block. At inference, the recurrent block can be applied a variable number of times. We pre-trained Dory1.0-4B on 6 trillion tokens. The first 4.5 trillion tokens were trained in NVFP4, and additional pre-training then extended the context length to 256K tokens. Next, we post-trained the model using Supervised Fine-Tuning (SFT) and Reinforcement Learning with Verifiable Rewards (RLVR), with a difficulty-aware loop curriculum that routes harder prompts to deeper loops. Increasing the loop count improves accuracy, with the largest gains on coding, agentic, and long-context benchmarks. With two or more loops, Dory1.0-4B outperforms larger models such as Qwen3.5-9B, Gemma 4 12B, and Gemma 4 26B-A4B on average across 12 benchmarks. Looping also makes Dory1.0-4B more token-efficient on agentic benchmarks: with four loops, it generates 1.3-2.5x fewer tokens than with one loop, as the extra computation is spent in latent space rather than on longer outputs. To reduce memory during looping, Dory1.0-4B caches keys and values only from the last loop, keeping its KV-cache footprint equal to that of a non-looped model at any loop count while retaining most of the accuracy gains. We open-source the BF16, MXFP8, NVFP4, and GGUF quantized checkpoints.

★ HybridSpec: Throughput-Aware Speculative Decoding for Hybrid GDN-MoE Models

Haiyu Wang, Sai Qian Zhang, Zechun Liu, Steven Li — Wed, 04 Nov 2026 00:00:00 +0000

97% · Efficiency 97%, Evaluation 67%

In speculative decoding, parallel drafters such as DFlash make drafting cheap, so target verification dominates each round. A central choice is the verification length, the number of tokens that the target processes in each round. Verifying more draft tokens can accept more of them, but work after the first rejection is wasted. For hybrid target models that combine Gated DeltaNet (GDN) and mixture-of-experts (MoE) layers, verification time grows much faster with this length than for dense models. In a single-request profile of Qwen3.6-35B-A3B with DFlash, it grew by 71% from length 4 to 16, almost entirely in GDN and MoE layers, compared with 4%-6% for a dense Qwen3-4B target. Each GDN layer must update a recurrent state at every token and recover it at the end of the accepted prefix. The MoE cost depends on the experts that the tokens activate, which are known only once verification runs. We present HybridSpec, which selects the verification length on the GPU from draft-side acceptance estimates and measured target latency. Verification then runs this length plus one extra token up to the first MoE router, before any expert runs. There, a lightweight routing correction uses the observed expert assignments to decide whether to keep that token. The rest of verification runs only on the selected tokens. Block-parallel GDN kernels avoid storing a state after every token and reconstruct the accepted state of every GDN layer in one launch. In isolated benchmarks at verification lengths 4 to 16 with every draft token accepted, these kernels, including accepted-state reconstruction, are $1.54\times$-$4.04\times$ faster than the default kernels. With three GDN-MoE target models in single-request greedy decoding, HybridSpec is $1.75\times$-$1.88\times$ faster than autoregressive decoding and $1.30\times$-$1.33\times$ faster than DFlash.

★ Say Less, Mean Better: Training-Free Verbosity Control for Language Models

Mahmoud Alyosify, Mirna Imbabi, Eman Elkhamisi, Rahatara Ferdousi — Wed, 04 Nov 2026 00:00:00 +0000

97% · Efficiency 97%, Evaluation 53%

Large language models (LLMs) often generate responses longer than necessary, increasing inference latency and computational cost. While prompting can encourage conciseness, controlling output length without compromising response completeness remains challenging. Aggressive early stopping can reduce generation costs but risks terminating responses prematurely. We introduce MinimalLM, a training-free framework for reliable verbosity control that combines uncertainty-aware, length-dependent stop-token adjustment with structural completeness guardrails. MinimalLM separates the decision of when to stop from whether stopping is structurally appropriate, coordinating response-level instructions, decoding-time intervention, and completion checks without modifying model parameters or requiring auxiliary models. We evaluate MinimalLM across three open-source LLMs and 330 questions, producing 5,940 responses across baseline and ablation configurations. MinimalLM reduces generated tokens by 47.6%, accelerates individual responses by 19-68%, and eliminates structurally truncated outputs in the primary evaluation. Despite substantial token savings, it achieves an average quality score of 8.04/10, compared with 8.42/10 for unmodified models and 6.46/10 for length-matched truncation. Evaluations across diverse subject domains (such as healthcare, programming, law, and science) demonstrate robust efficiency gains, while ablations establish the complementary roles of decoding control and completeness constraints. Our findings highlight that efficient generation requires more than producing fewer tokens: it requires controlling when generation ends while preserving response integrity. MinimalLM offers a lightweight, training-free approach toward this goal.

★ Interpretable Early Termination of Web Navigation Agents via Closed Sequential Pattern Mining

Sergio Talavera, Magdalini Eirinaki — Wed, 04 Nov 2026 00:00:00 +0000

97% · Agents 97%, Efficiency 93%, Evaluation 83%

Web navigation agents built on open-weight language models fail often and fail late, with most of the tokens spent after the trajectory has become unrecoverable. Recent work has shown that early termination is a viable cost-reduction strategy, but existing approaches produce opaque termination decisions that do not support diagnosis or human oversight. In this work we apply closed sequential pattern mining (BIDE+) to K-step prefixes of agent execution traces to produce a small, human-reviewable failure-signature library that organizes into a four-category failure taxonomy with associated repair recommendations. We use a multi-variant evaluation protocol that explicitly removes contamination artifacts (zero-step crashes and trivially separable single-step successes) so that reported detection accuracy reflects behavioral discrimination rather than trace-length distributions. We evaluate our methodology on 1,544 MiniWoB++ traces collected from three open-weight language models, achieving the best F1 at K=3 on the Exclude-errors configuration across eight methods (0.717), recovering 88% of test-set failures within three steps, and supporting an early-termination policy that saves 26.3% of tokens at 92% precision.

★ DualWorldBench: Can Agents Plan Deliveries across Symbolic and Grounded Worlds?

Yang Chen, Bin Wen, Danyang Peng, Hong-Jie You, Zi-Qiao Shang, Lan-Zhe Guo — Wed, 04 Nov 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 35%

Recent multimodal agents have shown promising progress in embodied planning and web interaction. However, existing evaluations often rely on simplified tasks with limited combinatorial structure, leaving it unclear whether current agents can robustly integrate perception and reasoning for multi-goal planning under temporal, physical, and multimodal constraints. We introduce DualWorldBench, a unified benchmark for evaluating agent planning under controlled symbolic--temporal and grounded--topological complexity. Built around a shared pickup--delivery interface, DualWorldBench consists of three complementary components: TextWorld, which isolates symbolic planning under varying task compositions and release dynamics; SimWorld, which evaluates grounded path planning under controlled topological variation with text, image, or multimodal inputs; and DualWorld-VQA, a diagnostic suite spanning perception, scene understanding, and planning. Experiments reveal a common gap between local understanding and globally consistent planning. In TextWorld, agents struggle to derive optimal task sequences under interacting temporal constraints; in SimWorld, grounded observations are not reliably converted into feasible action sequences. DualWorld-VQA further confirms that perception and topology understanding do not necessarily translate into constrained planning. Together, these results identify the central challenge as composing symbolic or grounded evidence into globally feasible and efficient multi-goal plans. These findings establish DualWorldBench as a challenging testbed for agents with tighter perception--reasoning--planning integration. All datasets and code are publicly available at our project website https://dualworld.netlify.app/.

★ Coarse Semantic Blocks for Efficient Training-Free Sparse Attention

Sosuke Kobayashi, Shoichiro Yamaguchi, Seiya Tokui — Wed, 04 Nov 2026 00:00:00 +0000

97% · Efficiency 97%, Evaluation 31%

We introduce a training-free sparse attention method that builds a prompt-specific key index to accelerate both prefill and decoding. The method clusters the keys into large semantic blocks that group tokens from noncontiguous positions. Each query position selects blocks by scoring their centroids and attends to the original keys and values within them. Queries that select the same block are processed together, so prefill stays fast even though each query chooses its own blocks. Large blocks make centroid scoring and this grouped processing cheap, so each query can attend to more tokens in the same time. Larger blocks with a larger token budget can thus be both faster and more accurate on long-context tasks than smaller blocks with a smaller budget. We evaluate two hybrid LLMs, Qwen3.5-4B and 35B-A3B, on one NVIDIA H200. Including the cost of building and using the index, whole-model wall-clock speedups over dense attention are 2.3-2.4x for prefill and 1.7-2.0x for decoding at 512K tokens, and 2.9-3.0x for prefill at 1M. The 35B-A3B model stays close to dense attention on five RULER tasks through 1M, on LongBench-v2 overall, and on MRCR through 256K. The 4B model loses accuracy mainly on multi-key retrieval, yet outperforms dense Qwen3.5-2B, whose latency is similar at 512K, on all three benchmarks.

★ LoopVQ: Sub 1-bit KV Cache Quantization for Looped Transformers

Minsoo Cheong, Sungjoo Yoo — Wed, 04 Nov 2026 00:00:00 +0000

97% · Efficiency 97%, New model 31%

Looped transformers are a parameter-efficient alternative to standard transformers for language modeling. By reapplying the same layers across multiple loop steps, recent looped language models match the performance of non-looped models several times their size. While this architecture effectively reduces parameter memory, each loop step still stores its own keys and values, so the key-value (KV) cache grows linearly with the number of loop steps and becomes the main memory bottleneck. We observe that the KV cache of looped models is highly redundant along the loop axis and that the loop steps differ substantially in their sensitivity to quantization. Based on these observations, we propose $\textbf{LoopVQ}$, a KV cache vector quantization method for looped transformers that codes channels of multiple loop steps jointly, and fine-tunes the codebooks to match the output of the full-precision model. At 0.5 bits per element, LoopVQ outperforms existing KV cache vector quantization methods for standard LLMs on general, long chain-of-thought reasoning, and long-context retrieval tasks with Ouro and Huginn models. Its design is orthogonal to these methods, and combined with TaSQ, it retains much of the BF16 accuracy even at 0.25 bits per element. We also implement LoopVQ in SGLang, where it nearly halves the decode latency at a 32K-token context and raises the peak serving throughput by up to $4.3\times$ and $6.1\times$ on the 1.4B and 2.6B Ouro models compared to the BF16 baseline.

★ LOLPlay: Benchmarking Real-Time Visual Agents in League of Legends

Zhaojian Yu — Wed, 04 Nov 2026 00:00:00 +0000

96% · Evaluation 96%

Visual agents need to coordinate reasoning and action while their environment evolves during inference. We introduce LOLPlay, a benchmark in native \emph{League of Legends} with 36 champion-specific tasks spanning resource management, combat, cooperation, and objectives. Agents interact through screenshots and mouse--keyboard actions; automated initialization and 82 executable subgoal checks distinguish partial progress from completion. Across 324 task evaluations of three models at three reasoning-effort settings, GPT-6 Astra achieves the highest task success rate of 25.0\% at low and medium effort, and the highest subgoal success rate of 40.3\% at medium effort. However, high effort improves neither metric over low effort for any model. For Astra, longer requests leave fewer interaction rounds, suggesting a cost to additional reasoning in real-time control. Trajectories reveal incomplete subgoal composition and actions executing after the observed scene changes, motivating agents that combine visual reasoning with timely action.

★ Multi-layer Indexer: Overlapping Selection and Attention in DeepSeek Sparse Attention

Serdarcan Dilbaz, Arjun Krishna, Kiran Kamble, Daniel M. Bikel — Wed, 04 Nov 2026 00:00:00 +0000

96% · Efficiency 96%

Sparse attention cuts the cost of long-context inference by making the attention matrix sparse: each query attends only a subset of the sequence, chosen by a fixed pattern or, in recent production LLMs, learned per query. In DeepSeek Sparse Attention (DSA), the attention of DeepSeek-V3.2 and the GLM-5 family, every layer is preceded by a lightning indexer that scores the whole cache and keeps the k=2048 highest-scoring tokens for the layer to attend. At long context this indexer scan, not the sparse attention it feeds, is where the decode time goes. But adjacent layers select nearly the same tokens, so a layer can attend the selection of the previous indexer layer instead of its own. This removes the dependency: each indexer then runs concurrently with the attention of its layer. The resulting Multi-layer Indexer (MLI) needs no retraining and no new weights. Its default realization changes no kernel, running inside SGLang's production decode graph; an optional fused decode kernel accelerates it further on DeepSeek-V3.2. MLI composes with IndexCache (the cross-layer reuse GLM-5.2 ships as IndexShare): IndexCache drops the indexer from some layers, and MLI hides the scan on the rest. We serve DeepSeek-V3.2 and GLM-5 under IndexCache's protocol at retentions 1/2 to 1/32 (the fraction of layers that keep an indexer) and contexts 10K to 195K. At equal retention the composition decodes faster than IndexCache: up to 14.5% faster on DeepSeek-V3.2 and 9.8% on GLM-5 at 1/2, and 7.9% and 5.1% at 1/4, largest where IndexCache keeps the most indexers and growing with context. At retention 1/r it is faster still than IndexCache at 1/(2r) while keeping twice the fresh selections: at 128K it reaches 1.32x the unmodified model's decode throughput at 1/4, above IndexCache's 1.28x at 1/8. Its LongBench, RULER and LongBench v2 accuracy stays within a few percentage points of IndexCache's.

★ AgentAM: An AI Agent for Spatter- and Defect-Aware Process Optimization in Laser Powder Bed Fusion

Olabode T. Ajenifujah, Amir Barati Farimani — Wed, 04 Nov 2026 00:00:00 +0000

96% · Evaluation 96%, Agents 92%, Hallucination 44%

Predictive models for laser powder bed fusion (LPBF) are rarely used on the shop floor because querying them requires programming and model-specific knowledge. We present AgentAM, a large language model (LLM) agent that gives process engineers natural-language access to CFD-trained surrogates of spatter and melt pool geometry for SS316L through the Model Context Protocol (MCP). Polynomial surrogates trained on 279 single-track simulations predict melt pool dimensions ($R^2 \geq 0.95$) and spatter volume ($R^2 = 0.83$ on test data) at any power-velocity point in the trained range, and geometric criteria assign each point a geometry-derived defect regime (keyhole, lack of fusion, balling or stable) that matches the regime obtained from the simulated geometry in 93.4% of tracks under cross-validation. The agent selects and parameterizes tools for prediction, process-window scans and constrained optimization. Minimum-spatter optima often sat within 10 $\mu$m of the lack-of-fusion boundary; a margin-aware optimizer moved them away from all thresholds and raised the estimated probability of remaining classified Good from 0.94 to 1.00 under an assumed 2% process drift, and from 0.72 to 0.91 when jointly sampled cross-validated geometry residuals were added. About 40-50 well-spread simulations sufficed to reproduce the defect map. To evaluate the agent layer, we introduce a 145-prompt benchmark scored against direct tool calls, plus eight prompts (40 runs) on margin-aware requests. Over the 725 runs of the main benchmark the agent passed the intended inputs in 98.5% of parsing prompts, reported the scored spatter values and defect classes faithfully and met every stated constraint in the 65 feasible optimization runs, within a median 1.0% of the brute-force optimum.

★ Going Deeper With Memories

Afshin Khadangi — Wed, 04 Nov 2026 00:00:00 +0000

96% · New model 96%, Efficiency 91%

Byte level language models avoid tokenizer dependence and provide universal coverage, but processing every byte at full resolution is expensive. We introduce DeepMem, a causal multiscale architecture that separates semantic compression from exact byte preservation. A learned invertible subband transform constructs a hierarchy of shorter sequences, while a lossless carrier and explicit detail sidebands preserve byte identity. Fast weight memory blocks operate along the encoder, with iterative shared state computation at the compressed bottleneck, while an attention based super resolution decoder restores byte resolution representations. DeepMem achieves stronger language modeling performance than a matched byte level Transformer on FineWeb-Edu and PG19, and obtains the best macro average performance across the evaluated FLORES languages. It also improves PIQA accuracy while remaining competitive on code modeling, multiple choice reasoning, and byte level corruption robustness. Bottleneck refinement requires approximately $7{,}600\times$ fewer counted FLOPs than equivalent full resolution reasoning, while maintaining competitive evaluation throughput.

★ The Geometry of Ignorance: Detecting Suppressed Knowledge in LLMs through Activation Fragmentation

Adar Peleg, Dvir Alsheich, Tomer Ashuach, Avi Mendelson — Wed, 04 Nov 2026 00:00:00 +0000

96% · Evaluation 96%

Evaluating whether a large language model (LLM) genuinely knows a concept remains a fundamental problem, since behavioral metrics can be fooled by suppression: a model that refuses to answer may still harbor recoverable knowledge. The framework measures the geometric coherence of \textit{anon-vectors}, the differences in hidden states between a concept-specific query and its anonymized paraphrase, and shows that genuine knowledge forms a tight, low-entropy cluster while ignorance forces geometric fragmentation. Across ten instruction-tuned LLMs from three model families, all fragmentation metrics reliably separate known from unknown concepts, and post-unlearning, fragmentation geometry correlates with adversarial recovery across four attack types and five unlearning algorithms, though the direction of that relationship is shown to depend on attack methodology. Furthermore, a layer-resolved analysis reveals that different adversarial mechanisms exploit this geometry at varying depths, with high attack-specific correlations that are invisible to behavioral evaluation. Together, these results constitute a single-phase, attack-free proxy for unlearning robustness that captures aspects of parametric knowledge inaccessible to output-level metrics. A reference implementation is available at https://github.com/AdarPeleg/geometry-of-ignorance

★ Alignment Faking Depends on the Value, Not Just the Model

Manodeep Ray, Vivek Srivastava, Shirish Karande, Murari Mandal — Wed, 04 Nov 2026 00:00:00 +0000

95% · Evaluation 95%

Language models can follow a directive when they believe they are monitored and abandon it when they believe they are not, a pattern termed Alignment Faking. Prior work establishes this for harmful-content objectives and a handful of moral-foundations values, leaving open how it distributes across a wider value space. We argue it is value-dependent: whether compliance survives a change in perceived oversight is a property of the model--value pairing, not of the model alone. Our framework profiles Alignment Faking per value, scoring each model-value pair under three conditions: bare (no system prompt), unmonitored, and monitored. On 32 ValueBench dimensions across ten open- and closed-weight models (8B-~675B), compliance gaps span 0.05-0.50, with a distilled 70B above a ~675B model. We observe that majority of the structural variance depends on which model was asked about which value and not on the model or value alone. A six-category taxonomy over reasoning traces tracks the behavioural shift: training-signal reasoning dominates under monitoring and value-conflict reasoning dominates when unmonitored. We package per-model results as value cards, extending model cards with behavioral disposition. Code and data are available at https://github.com/respailab/VALET.

★ VagueBench: Benchmarking the Robustness of Embodied Agents on Ambient Context Illusions

Yang Qu, Qiang Fu, Zonghao Ying, Jiayi Zhao, Yaoyuan Zhang, Jinyang Guo — Wed, 04 Nov 2026 00:00:00 +0000

95% · Evaluation 95%

Embodied agents are increasingly deployed in open-world environments, yet current robustness evaluations remain biased toward hazardous instructions and jailbreak vulnerabilities. A more fundamental issue is largely overlooked: agents are rarely tested under semantically sufficient but contextually minimal instructions (i.e., common in real-world human communication), where auxiliary contextual cues are absent. For the first, we identify a novel phenomenon termed Ambient Context Illusion, where agents fail to execute tasks correctly when non-essential contextual details are omitted, despite unchanged task semantics. To systematically study this, we establish VagueBench, a benchmark built upon the AI2-THOR interactive simulation environment for the evaluation of context dependency in embodied agents. We systematically consider key semantic components of instructions (entities, procedures, constraints, and prior knowledge) and derive a unified taxonomy with four categories and nine subtypes to reduce auxiliary information. VagueBench contains 1,303 instruction pairs across diverse indoor environments, constructed via a human-agent co-generation framework that integrates semantic redundancy estimation, taxonomy-guided transformation, and rigorous semantic-invariance verification. Through extensive evaluation on 6 state-of-the-art embodied agents, we observe significant degradation on VagueBench in both task performance and executability (relatively -19.65% in task success rate and relatively +54.10% in empty plan rate) revealing a strong reliance on auxiliary linguistic signals rather than core task semantics. Furthermore, we demonstrate that this phenomenon persists in real-world robotic systems, where it can lead to incorrect decisions and potential safety risks.

★ El Agente F\'armaco: A Scalable AI Scientist for Drug Discovery

Wed, 04 Nov 2026 00:00:00 +0000

95% · Agents 95%, Evaluation 87%, Efficiency 69%

Computational drug discovery offers a growing range of tools for proposing and evaluating molecules, but combining them and deciding what to do next still requires substantial expert effort. Here, we introduce El Agente F\'armaco, an AI scientist that uses modern large language model (LLM) technology to turn discovery goals expressed in everyday language into complex, scalable computational screening workflows. The agentic harness selects scientific tools, reviews their results, and adjusts subsequent steps to the target and the user's objectives. Our approach centers on a shared scientific memory: F\'armaco Atlas, which stores molecules, structures, and results together with their provenance -- a record of how they were obtained. The agent uses targeted database queries to retrieve the evidence needed for each decision. A single request can launch calculations for an entire collection of molecules, without requiring a separate LLM decision for every compound. This design limits the information the model must process as screening collections grow, while keeping the underlying results available for inspection and further analysis. The integrated tools support molecule generation, prediction of binding and molecular properties, and searches for compounds available for purchase. In a retrospective comparison with an expert-led campaign, the agent proposed chemically distinct candidates with competitive docking scores and higher calculated drug-likeness under explicitly imposed selection criteria. In a 22-task benchmark against plain tool calls and Codex, the agent achieved the highest observed mean model-judged score, with lower median cumulative token use and peak context. F\'armaco provides a foundation for flexible, scalable drug discovery workflows that connect scientific reasoning with an enduring, traceable record of computational work.

★ Thinking Fast, Trading Late: System One Decision Models vs the Market

Wed, 04 Nov 2026 00:00:00 +0000

95% · Evaluation 95%, New model 31%

Large language models (LLMs) read company news at scale, and backtests suggest that their readings predict stock returns. We ask how much of that predictability survives a test that rules out look-ahead, memorization and selection. We propose OLIVIA, a protocol for such tests: point-in-time inputs with company identities masked, selection on validation data only, an analysis plan frozen by cryptographic hash, and a test period unblinded once through a logged lock. We apply it to AI decision models, non-generative language models that return probabilities for typed questions, reading news about S&P 500 stocks: an open-weights model we fine-tuned on 2024-2025 data, three commercial models and standard baselines. The fine-tuned model's signal replicated on two validation quarters (daily information coefficients 0.055 and 0.065) but fell to 0.020 (t = 1.7) on the blinded January-July 2026 test. The commercial models' readings line up with the overnight move to the open (information coefficient 0.074 to 0.077) but not with the move after it, so the news is priced by the time a trader can act on it at the open. No model can tell over- from under-reaction, and no pre-registered strategy survives costs and a deflated Sharpe ratio. Commercial models identify masked companies in 62% to 74% of cases against a 28% baseline: masking hides names, not companies. The market reads the news first.

★ LoopKV: Loop-Aware KV Cache Compression for Serving Looped Transformers

Tianle Zhou, Yang Zhang, Tianyu Wu, Zhuohan Wang, William Nixon, Juncheng Yang — Wed, 04 Nov 2026 00:00:00 +0000

95% · Efficiency 95%, Evaluation 38%

Looped transformers apply a shared block several times per token to improve accuracy without adding parameters. For models such as Ouro that keep a separate key-value (KV) cache per loop, cache memory limits the batch size that amortizes repeated weight reads. Uniform-bit quantization and fixed rules per loop do not allocate precision according to each layer's sensitivity in each loop. Measurements on Ouro show that quantization sensitivity varies across (loop, layer) cells, and the sum of individual cell damages approximates the overall output damage. Sharing damage also varies by cell, with later loops generally tolerating sharing better. Skipping low-attention values to save reads produces errors that compound across loops. LoopKV uses a one-time offline profile to choose 4-bit storage, 2-bit storage or reuse of the final loop's KV for each cell. The planner minimizes reads under budgets for predicted damage and storage. During decoding, each loop reads the high 2 bits of each 4-bit value and fetches the low bits only where its own attention weights require 4-bit precision. A single fused kernel in vLLM executes the plan. Serving Ouro-1.4B on one RTX PRO 6000, the memory-optimized LoopKV-M plan stores 35% less KV than uniform 4-bit quantization (KIVI-4) on the same fused kernel and CUDA graph. It sustains 1.7-2.2x KIVI-4's goodput under a latency service-level objective with a 16K-token maximum length. None of its averages on six public benchmarks falls significantly below BF16's.

★ Anchor: Augmented Cross-Layer Recurrent State for Long-Context Retrieval in Test-Time Training

Vimal William, Yuanzhe Li, Danella Zhao, Huanrui Yang — Wed, 04 Nov 2026 00:00:00 +0000

95% · Efficiency 95%, New model 83%

Parameter-efficient linearization turns a pretrained autoregressive Transformer into an efficient model by hybridizing sliding-window attention (SWA) with a constant-size recurrent memory. But does every layer's recurrent state contribute to long-context retrieval? Investigating an SWA and Test-Time Training (TTT) hybrid, we find that retrieval is sparse across depth, concentrated in a few anchor layers, and that the remaining states hold similar information expressed through layer-specific transformations. We exploit both observations in Anchor: TTT memory placed only at the anchor layers, and a cross-layer recurrent state written once and read by many layers, each through its own learned read map that realizes a layer-specific stretch and rotation, along with demonstrated generalization to other hybrid attentions. On Llama-3.1-8B and Mistral-7B, Anchor delivers 2.3-3.1x higher recall and 6.7-8.9 points higher MMLU accuracy than LoLCATs with a 10x smaller recurrent state, retains 91-95% of the teacher's MMLU accuracy, and cuts end-to-end latency by up to 4.3x relative to softmax attention.

★ Self-Privileged Fine-Tuning

Yangrong Hu, Xueyu Zhou, Yue Wu, Jian Huang — Wed, 04 Nov 2026 00:00:00 +0000

95% · New model 95%

Supervised fine-tuning (SFT) is a key component of language model post-training, enabling effective task adaptation and serving as an important foundation for subsequent reinforcement learning. Recent methods seek to improve SFT by introducing token-level weights into the training objective, where these weights are typically derived from a model's output distribution at each token position. To better understand how such weighting schemes affect SFT, we analyze their optimization objectives and theoretically characterize the target distributions they induce. Our analysis shows that dynamic weighting methods, such as Dynamic Fine-Tuning (DFT), induce a heavily sharpened distribution, whereas static weighting methods fit a target distribution proportional to the product of the data distribution and a frozen prior. These findings suggest that static weighting provides a more stable training target, while its effectiveness critically depends on the quality of the prior distribution. Building on this observation, we propose Self-Privileged Fine-Tuning (SPFT), which uses a frozen copy of the initial model conditioned on privileged information (such as ground truth for mathematical questions) as a prior to provide guidance for token-level weighting during fine-tuning. Through a Bayesian decomposition, we further show that SPFT reweights the standard static prior according to how informative each target token is for predicting the ground-truth outcome, thereby assigning larger weights to more predictive tokens. Experiments on mathematical reasoning and code generation show that SPFT improves over standard SFT and other baselines across the main settings.

★ Think Less, Locally: Local Preference Learning for Reasoning Compression

Muyao Yuan, Liyan Xu, Weizhan Zhang, Muyan Jiao, Jiangnan Li, Jinchao Zhang — Wed, 04 Nov 2026 00:00:00 +0000

95% · Efficiency 95%, New model 94%, Evaluation 46%

Large reasoning models achieve strong performance on challenging tasks, but their long chain-of-thought often incurs substantial inference costs by continuing to deliberate after sufficient information has been obtained. Existing reasoning compression methods primarily operate on reasoning trajectories, paying less attention to the local decision of whether further reasoning is necessary given the current state. We propose LoCo, a local preference learning method for reasoning compression. LoCo detects candidate overthinking states from the model's own traces, freezes the corresponding reasoning prefixes, regenerates shorter verified continuations from the same states, and trains on the resulting state-conditioned preference pairs with DPO. This formulation teaches the model when to continue reasoning and when to conclude, rather than simply encouraging shorter trajectories overall. We construct training data entirely from self-generated mathematical traces and evaluate LoCo on five mathematical and three code generation benchmarks. LoCo reduces completion length by 41.7\%--45.4\% on mathematics and 23.4\%--36.3\% on code, while maintaining or improving accuracy and achieving the best AUC across all code benchmarks. LoCo consistently improves the accuracy-efficiency trade-off on mathematics and transfers effectively to code generation despite math-only training, demonstrating the cross-domain generality of local preference learning for reasoning compression. The code and model will be publicly released at https://github.com/MuyaoYuan/LoCo.

★ Who is the Pronoun Pro? Comparing LLMs and Humans at Number-Disambiguation of 'they'

Carlos Hartmann — Wed, 04 Nov 2026 00:00:00 +0000

95% · Evaluation 95%, Efficiency 63%

This study evaluates whether large language models (LLMs) can reliably replace or complement human research assistants in linguistically complex annotation. In a controlled annotation task, 73 human annotators (local students and crowdworkers with and without a background in linguistics) and more than 40 LLM configurations number-disambiguated 174 authentic uses of 'they' from Reddit, including innovative and context-sensitive cases. The best LLM, Kimi K3, reached 94.5% accuracy, above the human majority vote (86.2%) and the mean of the top half of human annotators (78.5%); only the single best human annotator scored higher (95.7%). LLMs also offer substantial practical advantages in cost and speed. However, qualitative error analysis reveals a crucial complementarity: human reasoning in low-agreement cases exposes inconsistencies in the annotation typology and thereby improves the methodological framework itself. The study contributes (i) a benchmark-style human-LLM comparison for a semantically demanding task, (ii) transparent cost and time evidence, and (iii) a methodological argument against a purely replacement-oriented view of automation. LLMs are viable for large-scale annotation support in linguistics, but their responsible use requires human critical oversight. The dataset with its gold-standard labels, all human and LLM annotations, and the fully reproducible code are openly available as a Quarto-based, public repository.

★ From AutoResearch to Continual AutoResearch: Do Agents Learn Across Tasks?

Jeremy Qin, Ruizhi Deng, Mohamed Osama Ahmed, Frederick Tung, Leo Feng — Wed, 04 Nov 2026 00:00:00 +0000

95% · Evaluation 95%, Agents 75%

Autoresearch has shown that agents can autonomously write code, run experiments, and iteratively improve machine learning solutions. Yet most evaluations start a new agent session from scratch for each new problem, making them costly and leaving unclear whether agents can retain and apply prior insights when adapting to new tasks. We introduce CONTINUAL AUTORESEARCH, an open-ended environment that compares persistent and reset instances of the same agent under matched task streams, data, and time budget. Continual AutoResearch covers nine task families, spanning synthetic tasks, symbolic regression, time-series forecasting, and tabular AutoML. Across five frontier agents, we measure first-solution quality, research efficiency, and final held-out performance. To distinguish insight transfer from session continuity, we run a memory ablation that vary what information agents retain from prior tasks. We show that persistent state improves research efficiency of all five agents, allowing them to find stronger solutions sooner. However, these gains do not consistently translate into better final solutions. Overall, our findings suggest that retaining experience and utilizing it effectively are distinct capabilities.

Honest Small Decision Models for Finance: Scale, Specialization Collapse, and Mixed-Corpus Remediation

Ajitabh Kumar — Wed, 04 Nov 2026 00:00:00 +0000

94% · Evaluation 94%, New model 57%, Hallucination 44%

Financial workflows need fast, calibrated, auditable decisions, not just generated text. We train System-1 decision models, small language models fine-tuned to answer typed questions (choice/score/noul) with a probability in a single forward pass, across two model families (K2-Horizon and Qwen3.5/3.8) and nine rungs from 0.8B to 32B parameters, using one frozen LoRA recipe on a public decision benchmark and two new finance corpora: a programmatic-gold policy corpus (rule-reading with exact labels) and a reality-labeled outcome corpus (19,640 realized forward-return questions in walk-forward splits, plus a sealed 2026 test of 1,360 questions). Narrow fine-tuning costs general decision competence 9-45 percentage points, a data-mixture effect that shrinks with model scale and is associated with narrow-gradient volume. On reality-labeled questions, every trained rung from 0.8B to 32B lands at the horizon-conditioned prior baseline (accuracy 0.37-0.38, no detectable conditional edge under our controls); the models respond with flat, well-calibrated probabilities that track per-horizon base rates: naturally honest uncertainty, with a mid-scale calibration sweet spot (ECE 0.026 at 2B, degrading to 0.043-0.044 at 9-32B under prior drift). A mixed-corpus remediation carries all competencies in one adapter at modest cost: it matches or slightly exceeds the generalist specialist on the public benchmark while holding the policy ceiling and near-specialist outcome calibration, yielding a deployable artifact with per-domain temperature regimes and a band-dispatch escalation analysis: sharp, dispatchable bands where rules determine answers; flat, review-routing bands where reality does not.

What Is This Memory Useful For? Measuring Task-Conditioned Utility in Long-Term Agent Memory

Chih-Hsuan Yang, Anjir Ahmed Chowdhury, Feng Yan, Rajeev Thakur — Wed, 04 Nov 2026 00:00:00 +0000

94% · Evaluation 94%, Agents 85%

A stored experience may answer one future question and offer little help with another. We measure this task-conditioned utility rather than assigning each memory a single importance score. A task-blind writer stores a memory without seeing the later task; a frozen reader then answers that task with and without the memory, using identical filler. A cost ledger records the resources used to form, store and expose the memory. On single-call factual-recall tasks built from LongMemEval-V2 web-agent memories, the same memory helps more when it contains information needed for the answer. Its mean paired reward gain on these answer-bearing task pairs exceeds that on matched-control pairs by $+0.33$ (95% interval $+0.21$ to $+0.46$). The tested task-blind summaries and structured records yield smaller gains than Raw, including at fixed total context; a budget sweep is consistent with retention of answer-bearing content contributing to the difference. The bearing--control contrast remains positive in a new session with randomized memory position and under a second reader, although individual memory--task labels change. Generic retrieval performs above chance but misses most screen-defined answer-bearing memories. The factual-recall gains do not extend to the tested software-engineering and scientific-coding benchmarks, where even raw memories have no measurable effect. We provide a public, versioned Conditional Memory Utility Dataset of paired measurements and costs. The practical takeaway is to evaluate memory compression by the task utility it preserves alongside its costs. Task-conditioned measurements make this comparison possible; their stability must also be checked.

P$^2$-MOPD: Accelerating Multi-Teacher On-Policy Distillation via Parameter Priors

Sinuo Wang, Jiacheng Lu, Zeyu Zhang, Maojia Song, Rui Sun, Beidi Luan — Wed, 04 Nov 2026 00:00:00 +0000

94% · New model 94%

Multi-teacher on-policy distillation consolidates expert capabilities by supervising student-generated trajectories, whereas task arithmetic offers a training-free alternative through linear parameter combinations. We view the linear combination of student and expert parameters as a single differentiable model and propose P$^2$-MOPD, which jointly optimizes the mixing coefficients and the student weights under the same on-policy distillation objective. Unlike task arithmetic, which stays within the span of the frozen student and expert parameters for any choice of merging coefficients, P$^2$-MOPD keeps the student trainable, so the experts act as parameter priors without confining it to this span. Across four domains at 1.7B and 4B scales, P$^2$-MOPD achieves faster and more balanced recovery than MOPD from the same initialization, surpassing its final average recovery in half the training steps. At the end of training, our 1.7B model reaches 94.7% average recovery, versus 80.9% for MOPD.

Evaluation Contracts and Evaluation Leverage: Sliced-Gap Readings in Structured Prediction Can Move Without Model Change

Hao Lu, Weixing Zhu, Ruiyang Li — Wed, 04 Nov 2026 00:00:00 +0000

94% · Evaluation 94%

NLP evaluation reports routinely cite sliced performance differences, most prominently the intra-inter gap in document-level relation extraction, as capability evidence. We show that such gap readings are joint properties of models and evaluation contracts: scoring function, gold version, threshold, output format, and prompt language. With predictions frozen, a method ordering reverses under a gold-standard exchange (+0.0363, CI excluding zero; tie-level after), movement localizing where gold re-partitions: renames are gauge transformations (Delta=0; re-annotation +0.862). We formalize this as an evaluation leverage account centered on the rank-flip condition. Constructively, the audited evaluation reading (AER) is a contract-marginalized, tie-declared, seed-prescribed, gate-filtered procedure whose tie declarations cover every contract-induced flip on the declared WMT19 grids (14/14; held-out 73% vs. union-of-ties 89%), matching the best single metric contract at epsilon=0.10 equivalence. Evidence spans DocRE, HacRED, a UD probe, and released gold revisions; an alternative-reference-set replication extends it beyond our testbeds. Failed predictions retained as boundaries; a 5% violation gate flags unverifiable settings and bounds survivor-selection bias. We release EvalLever, a preregistered workflow with an amendment trail; on LLM-as-judge grids, the audited ensemble stabilizes every unstable pair (13-15 to zero; generic ensemble shrinkage: the preregistered correlation criterion failed, the companion stabilization criterion passed).

PACT: Certifying What Is Written in Parallel Decoding of Diffusion Language Models

Yanchao Li, Yusu Zhang, Jiaqing Xie, Yi Ban, Ben Gao, Zhongyu Huang — Wed, 04 Nov 2026 00:00:00 +0000

94% · Efficiency 94%, Evaluation 51%

Parallel decoding speeds up masked diffusion language models by writing several tokens per forward pass. Published commit rules choose these tokens by testing the predicted distributions. Under sampling the written tokens are drawn, and these tests do not see them. We find that answer quality follows the reliability of the tokens actually written. An answer's log probability splits into this reliability and the dependence within each step. We propose PACT, a training-free commit rule that draws the tokens first, certifies the set it will write and spaces the positions of each step apart. This certificate holds at every temperature, and we show that no rule with the same guarantee commits more positions. Experiments use four models of the LLaDA and Dream families and HumanEval, MBPP, GSM8K, MATH500 and IFEval benchmarks. PACT reaches the quality of published commit rules with fewer forward passes. Spacing removes positions from each step, yet it usually lowers the total cost because later steps then write more tokens. At the same speed, published decoders also write a word twice in a row far more often than PACT. These results suggest that commit rules should certify the tokens they write. Code is available at https://github.com/LiZaiyuan0619/PACT.

COMPREHEND: Ultra-Fine-Grained Benchmark Understanding by Entailment Hierarchy

Wed, 04 Nov 2026 00:00:00 +0000

94% · Evaluation 94%, Hallucination 36%

Understanding what kind of prompts users typically ask and how well an LLM performs on these prompt categories is crucial for LLM practitioners. The existing methods satisfy the demand by hierarchically clustering the user prompts and synthesizing the cluster names using an LLM to build a taxonomy. However, the cluster names from these methods are usually too general for humans to understand the detailed prompt distributions and identify the exact strengths and weaknesses of LLMs. In this study, we propose COMPREHEND, which automatically constructs an ultra-fine-grained entailment hierarchy for the prompts. The entailment hierarchy replaces the cluster names in the taxonomy with sentences to provide more precise cluster descriptions for more fine-grained categorization. Compared with EvalTree, a state-of-the-art taxonomy construction method, COMPREHEND improves an entailment/faithfulness metric from 22.6 to 41.7 and an informative metric from 16.3 to 39.1 in WildBench. Our experiments also show that the weakness report generated based on our entailment hierarchy is much more informative than EvalTree in terms of predicting the response scores in the benchmark. All the code is released in https://github.com/JackNapier20/COMPREHEND_public.

Bidirectional Centered Score Matching for Discrete Diffusion Distillation

Peng Tu, Baoyou Chen, Haoyuan Xia, Zilong Dong, Weihao Yuan, Siyu Zhu — Wed, 04 Nov 2026 00:00:00 +0000

94% · Efficiency 94%, New model 90%

Discrete diffusion models with parallel decoding are often distilled using token-wise KL objectives. Although token-wise KL exactly matches the corresponding product-form block distributions, its probability-space optimization geometry and teacher-only candidate truncation may provide insufficient supervision under aggressive few-step distillation. Concrete Score Matching (CSM) provides an alternative within-position objective by aligning relative logit geometry, but its complete pairwise form incurs $\mathcal{O}(|\mathcal{V}|^2)$ complexity. We propose Bidirectional Centered Score Matching (BCSM), an efficient within-position distillation objective that constructs an $\mathcal{O}(K)$ bidirectional active set, reduces pairwise comparisons to $\mathcal{O}(K^2)$, and factorizes them into an $\mathcal{O}(K)$ objective. We further introduce cross-timestep transition alignment. Experiments demonstrate strong quality--efficiency trade-offs, including up to $6\times$ NFE reduction with a 16.9\% relative HumanEval gain and a 15.1\% relative MMMU gain at single-step inference.

Training Specialist LLMs on Your Phone: Forward-Only Optimization via Block-Spectral Coordinates

Liyan Tan, Yequan Zhao, Ziyue Liu, Yupeng Su, Ruijie Zhang, Zheng Zhang — Wed, 04 Nov 2026 00:00:00 +0000

94% · New model 94%, Evaluation 49%, Efficiency 48%

Large language models (LLMs) on personal devices keep user data local, but limited memory restricts their general capability. Since most user-specific tasks do not require the breadth of frontier models, a practical solution is to adapt a shared on-device model into lightweight specialists. However, backpropagation is prohibitive in memory, and users rarely provide labeled data. We present MobileSpec, a system that trains specialists with forward passes alone and learns from automatically checked task outcomes, as in reinforcement learning with verifiable rewards (RLVR). Because forward-only updates are built from random perturbations, training is efficient only when these perturbations resemble the weight changes of RLVR, which largely preserve the singular values of pretrained weights and rotate their singular vectors. The core algorithm of MobileSpec, FuRA-S, produces such perturbations while training only the block-spectral coordinates of pretrained weights, namely the singular values of their column blocks. As we show theoretically and empirically, rescaling these blocks independently shifts their relative contributions to each weight matrix and thereby rotates its singular vectors. Since this update can be rewritten as a transformation of the layer input, FuRA-S also adapts INT4 models without modifying their weights. With specialists of only 1-5 MB that update 0.015-0.05% of the parameters, MobileSpec raises the success rate of a phone-use agent from 6-23% up to 89-100% and trains a personalized 8B model on an iPhone 16 Pro overnight. On RLVR benchmarks, FuRA-S is competitive with full-parameter evolution strategies (ES) and outperforms ES over LoRA adapters and quantized ES on INT4 models.

ARLoop: Fully-Open Pre-trained Looped Transformers with Attention Residuals

Yuekun Yao, Yupei Du, Dawei Zhu, Huan Sun — Wed, 04 Nov 2026 00:00:00 +0000

94% · New model 94%, Efficiency 50%

Looped transformers increase effective depth by repeatedly applying a shared set of transformer layers, offering a parameter-efficient way to scale computation. However, standard residual connections aggregate information across recurrent depth into a single stream, providing limited control over which intermediate representations from preceding layers should be reused. In this paper, we introduce ARLoop, a looped transformer architecture based on attention residuals, where the input of each layer is selectively aggregated through attention over model depth. Distinct from standard transformers, we find that looped transformers do not require access to the full residual history: sliding-window depth attention matches or outperforms full-history attention at larger recurrent depths. Across controlled experiments with multiple parameter scales (140M and 370M) and recurrent depths (R = 1 to 5), ARLoop consistently outperforms prior residual designs for looped transformers. More importantly, we further scale ARLoop to 0.6B and 1.2B parameters under a realistic LLM pretraining recipe, demonstrating that the gains transfer to realistic pretraining, with particularly strong improvements on math and coding tasks. Our 1.2B model, pretrained on only 0.5T tokens with three recurrences, outperforms Huginn-3.5B (the only fully-open looped LLM as of our release date) across all evaluated tasks and is competitive with similar-sized non-looped models such as Qwen3.5-2B despite being trained on substantially less data.

Order-Only On-Policy Distillation

Guangzeng Han, Yu Wu, Xiaolei Huang — Wed, 04 Nov 2026 00:00:00 +0000

94% · New model 94%

On-policy distillation (OPD) trains a student on its own generations by matching the teacher's token probabilities. These probabilities entangle two kinds of information: which tokens the teacher prefers, and how confident it is. The latter reflects the teacher's own capacity and calibration, which a smaller student may be unable to reproduce. Because the teacher temperature changes these values without changing the ordering, OPD is also sensitive to the temperature choice. We ask whether the teacher's ordering of candidate tokens is enough, and propose Order-Only On-Policy Distillation (O^{3}PD). At each student-generated prefix, O^{3}PD trains the student to reproduce the order of the teacher's top-K tokens, without fitting their probability values. The objective is therefore invariant to the teacher temperature and needs no temperature tuning. Across six teacher--student pairs, O^{3}PD outperforms OPD on the average over HMMT25, AIME24, and AIME25 in every pair, by up to 2.9 points, with gains of up to 5.6 points on AIME25. Other ranking losses yield similar gains, so the benefit comes from ordinal supervision itself. Analysis shows that O^{3}PD departs further from the teacher's probabilities than OPD, yet reproduces its ranking more faithfully and yields a stronger student. The ranking, not the probability values, thus carries the knowledge that on-policy distillation needs to transfer.

Stress-Testing the Robustness of LLM-Based Causal Inference

Wed, 04 Nov 2026 00:00:00 +0000

93% · Evaluation 93%, Hallucination 42%

Why do some of the world's most powerful Artificial Intelligence (AI) systems still struggle to reason about cause and effect? Causal discovery, the ability to discern the underlying drivers of a system, is fundamental to scientific inquiry and has traditionally relied on structured data and statistical methods. Recently, Large Language Models (LLMs) have been explored as an alternative paradigm for causal inference, leveraging their broad knowledge and reasoning capabilities. Yet, despite their linguistic fluency, LLMs often rely on surface-level pattern recognition rather than genuine causal logic. While it is known that LLMs stumble in this domain, we lack a precise map of where, why, and how significantly these failures occur. In this paper, we propose a systematic stress-test framework across six critical dimensions: model scale, causal graph complexity, prompting strategy, metadata quality, data quality, and sensitivity to uncertainty. Our findings reveal a significant "optimism bias" in LLMs; models consistently overestimate causal links, leading to low precision across the board. Furthermore, we demonstrate that traditional evaluation metrics can be deceptive: while reasoning-heavy models lead in performance, the highest scores often occur when uncertainty in inference is overlooked. Once uncertainty is integrated into the evaluation, performance drops, providing a more honest reflection of causal reliability. We also find that, while scaling model size improves outcomes, these gains are fragile; performance degrades sharply as the quality of metadata decreases. This work serves as a practical guide to the blind spots of LLMs, offering a clear-eyed assessment of when these tools can be integrated into a causal discovery pipeline and when they are likely to lead researchers astray.

Quantifying out-of-distribution data severity with Wasserstein distances: benchmarking large language models for alloy property prediction

YuJie Liu, Dezhen Xue, Jun Sun, Xiangdong Ding, Zhibin Gao, Turab Lookman — Wed, 04 Nov 2026 00:00:00 +0000

93% · Evaluation 93%, Efficiency 51%, New model 45%

As general-purpose Large Language Models (LLMs), such as Gemini and DeepSeek, exhibit increasingly advanced reasoning capabilities, a critical question arises: how do their property prediction abilities compare to those of domain-specific pretrained materials models? In this study, we evaluate predictions for Ti, steels, high-entropy alloy (HEA) and Al, demonstrating that general-purpose LLMs surprisingly outperform SteelBERT, MatSciBERT, and SciBERT. To further enhance performance, we introduce a hybrid approach that utilizes numerical priors from a prior model, such as CatBoost or SteelBERT, to guide the general LLMs (Deepseek V3.1, Gemini-2.5-Flash). By inputting these initial estimates from a prior model as prompts, the LLM acts as a reasoning corrector, achieving accuracy superior to the prior model and "unguided mode". Furthermore, we investigate the impact of data strategy using a similarity-based metric, revealing that data quality dominates quantity; using merely 5% of high-similarity reference data of training set as input yields performance similar to the results of using 100% of training set. Additionally, the LLMs demonstrate excellent uncertainty calibration, where self-reported "high" confidence strongly correlates with minimal deviation from ground truth. Our findings suggest that synergizing general-purpose LLMs with prior model numerical baselines and high-quality data curation creates a cost-effective and accurate paradigm for materials discovery, reducing the need for expensive domain-specific pretraining.

Performance, Scalability and Maintenance: An Experimental Study of RPA vs. Autonomous Agents in the Healthcare Domain

Wed, 04 Nov 2026 00:00:00 +0000

93% · Agents 93%, Evaluation 36%

Healthcare information systems are fragmented, and retrieving data from legacy web portals is a recurrent need. Robotic Process Automation (RPA) is deterministic but sensitive to interface changes, whereas autonomous agents based on large language models (Agent Process Automation, APA) promise adaptability at a higher computational cost. Following Design Science Research, we developed and compared an RPA solution (Playwright) and an APA solution (browser-use, LangGraph and Gemini 2.5 Flash-lite) to retrieve mammography and breast ultrasound reports from a laboratory portal, treating reports solely as informational artifacts. Across two rounds (T and T+7), RPA reached 98.90% accuracy and 98.27% F1-score versus 96.49% and 94.48% for APA, was about 13 times faster (16.07 s vs. 207.21 s per patient) and cost about six times less per report (USD 0.030 vs. 0.187). Both reached 100% step success rates; APA varied more over time. For a stable interface, RPA was superior.

ROCm InferaSim: Cut LLM Serving Trial Costs and Deploy the Best Configuration

Anshu Raina, Yao Fu, Zhang Jiejing, Peyman Razaghi, Shiksha Patel, Ravi Gupta — Wed, 04 Nov 2026 00:00:00 +0000

93% · Evaluation 93%, Efficiency 91%

Serving an LLM efficiently depends on the engine, how the model is split across GPUs, precision, scheduler settings, speculative decoding, KV-cache management, and prefill-decode disaggregation. The best configuration depends on the model, hardware, workload, and objective. Operators usually find it by trial and error. Exhaustively measuring the candidates in our study would take about 25,000 GPU-hours. ROCm InferaSim predicts serving performance from a few short measurements. Choices such as batch size, prompt length, and GPU count change the amount of work per GPU without changing the kernels. ROCm InferaSim benchmarks each distinct set of kernels once, on at most one node. We call this measurement an anchor. Analytical laws carry that anchor to other deployments that run the same kernels. For multi-node deployments, ROCm InferaSim reduces the configuration to a single-node equivalent, measures it, and scales the result analytically. A simulator replays multi-turn traffic, and a tuning agent searches configurations using the predictions. On AMD Instinct MI355X GPUs with vLLM, SGLang, and ATOM, 11 anchors took 9.1 GPU-hours, against 204 GPU-hours to measure the 226 fixed-length configurations they cover. On 172 configurations, time per output token has 8.1% median error. On a multi-turn coding workload, throughput is within 8%. For 159 disaggregated deployments, none used as an anchor, throughput is within 5.5%. A five-candidate short list contains a deployment within 12% of the best feasible pairing of model and latency target.

Where Did the Update Go? Certifying Declared Change Envelopes for Model Updates

Muhammad Usama, Muhammad Saad Qadir — Wed, 04 Nov 2026 00:00:00 +0000

93% · Evaluation 93%

Deployed language models are updated continually, and regulators increasingly accept changes that were declared in advance. Showing that an update changed behavior only where it was declared to is a different statistical problem from bounding how much it changed: a useful update is meant to change some decisions, so global disagreement and regression bounds cannot certify it. We formalize the problem with declared change envelopes. Before training, the developer commits to a region of inputs, and every decision that changes outside it counts as leakage. Leakage compares two models, so it can be certified from unlabeled traffic with anytime-valid tests whose sample cost we bound from above and below, and per-update certificates compose into a bound on lifetime drift outside all declared scopes. We measure leakage for LoRA and full fine-tuning, preference optimization, reinforcement learning, and knowledge editing on models from 1.5B to 72B parameters. Narrow updates leak widely: domain and policy LoRA updates keep only about 7-36% of the decisions they change inside the declared scope. Across 256 trained variants, leakage tracks the update's KL divergence to the incumbent on out-of-scope inputs (Spearman 0.94). None of the customary remedies we tested makes a rule-like update certifiable without giving up much of its intended change. Distilling the incumbent's decisions on unlabeled prompts in the deployment format does: anchored LoRA, GRPO, and (at a small learning rate) full fine-tuning updates certify 5% leakage from a few hundred to a few thousand unlabeled samples, and the certificates replicate on held-out traffic and on written answers. The certificates cover decision-like outputs only; on open-ended chat, re-batching alone changes 11-21% of responses and no update certifies. Code: https://github.com/Usama1002/declared-change-envelopes

Sparse Internal Control of Language Models

Xizhe Zhang — Wed, 04 Nov 2026 00:00:00 +0000

93% · Evaluation 93%

Language models are increasingly used in text generation, decision support, and automated interaction, creating a need for localized, reversible, and selective control of their behavior. Activation steering demonstrates that internal interventions can alter the behavior of frozen models, but how to select and evaluate compact, multi-site control interfaces remains an open question. I introduce sparse internal control, a framework for steering target behavior by applying inference-time interventions to a small set of internal model nodes. A local target-controllability formulation characterizes candidate site--direction pairs by their behavioral effects and specifies constraints on target recovery, off-target preservation, and intervention energy. Two complementary selectors instantiate this formulation: Driver-OMP selects nodes for sparse reconstruction of a desired behavioral displacement, while Coverage-Driver favors stable coverage across prompts. I propose a six-axis control-evaluation protocol measuring reachability, signed reversal, dose response, feedback controllability, off-target preservation, and held-out reuse. Exploratory experiments on indirect object identification, MMLU multiple-choice question answering, sentiment steering, and refusal steering provide evidence for the feasibility of steering task-defined behavior through sparse internal interventions. The six-axis evaluation characterizes these interventions beyond target movement, revealing distinct control properties that depend on the task and available direction information. Together, the formulation, reference algorithms, and evaluation protocol provide a foundation for the systematic study and development of internal control methods for language models.

CogniLife: Continuous Cognitive Distillation for Deployed Multi-Expert Agent Marketplaces

Xuwei Tan, Jingchun Huang, Ming Cai, Shuangquan Lyu, Tiange Wang, Yinzhi Wang — Wed, 04 Nov 2026 00:00:00 +0000

93% · Evaluation 93%, New model 33%

Multi-expert agent marketplaces -- where creators publish domain-specific AI experts for end users to consult -- are an emerging deployment pattern for language technology. A published expert can receive many requests yet little evidence about its creator, who remains active elsewhere: the cold-expert problem. We frame continuous cognitive distillation as converting attributed creator observations into revisable decision rules, communication patterns, and domain procedures. CogniLife proposes a cross-expert attribution design for this problem, with Creator Profiles for transferable creator patterns and Expert Journals for listing-specific knowledge; this design is not yet implemented or evaluated. A deployed platform partially realizes the surrounding lifecycle through creator-authorized authoring, capability-owned proposal filtering, and reversible consolidation of eligible owner-scoped memory. On ProfyBench-v1, an 85-task offline benchmark built from replay seeds and controlled internal tests, the Profy Full configuration as a whole achieves a mean overall score of 4.65/5 from the primary LLM judge, compared with 4.01 for Raw LLM and 4.28 for Expert Prompt, with the largest gains in completeness, specificity, and missing-information tasks. These configuration-level scores do not test cold-expert mitigation, cross-expert transfer, or longitudinal distillation.

MIMIC-Note-Extract: A Large-Scale Structured Dataset Extracted from MIMIC-IV-Note Discharge Summaries

Wed, 04 Nov 2026 00:00:00 +0000

93% · Evaluation 93%, Hallucination 33%

MIMIC is one of the most widely used public hospital datasets for machine learning on electronic health records, yet clinically important information, such as past medical history, allergies, and diagnostic findings, remains confined to free-text discharge summaries. Studies requiring this information must therefore build their own extraction pipelines, duplicating substantial effort across projects. We release MIMIC-Note-Extract (MNE), a large-scale dataset that uses a hybrid extraction pipeline combining deterministic parsing with schema-guided large language model generation to transform 16 sections from 300,720 discharge summaries into structured tables. MNE adds information that has no structured MIMIC counterpart and enriches related tables with broader coverage and greater clinical detail, including 10 times more patient-history entries and medication histories for more than twice as many admissions. Validation by medical doctors confirms extraction accuracy above 92% across sections. Using MNE, we construct a clinical decision-making benchmark for LLM agents, demonstrating how reusable structured representations of clinical narratives reduce data curation effort.

Residual Attribute Leakage Is an Interface: Auditing Privacy Evaluation After Text De-identification

Tuan Thanh Duong, Joongmin Shin, Jeongho Yoon, Junyoung Son, Heuiseok Lim — Wed, 04 Nov 2026 00:00:00 +0000

93% · Evaluation 93%

A privacy evaluation of de-identified text asks a language model (the adversary) to guess a person's protected attribute. We introduce a prior-normalised clipped leakage score of its confidence gain on the true value and audit the conclusions drawn from it. Each verdict depends on an evaluation interface (the queried person, answer options, prior, adversary model and scoring rule), which we vary on fixed releases of three corpora. In a stress test, an editor that writes the opposite-sex form at the same spans a redaction editor would redact misleads the adversary about most people, yet ranks first among editors scored on the same MedQA entries under this score, as under attack accuracy and proper scoring rules read as privacy. A relabelling-invariant reading of AUC ranks that editor sixth of eight, and a rule-aware adversary recovers the attribute from the same text. On ordinary releases most verdict changes are gains or losses of separability; of six choices varied over all pairs, only the adversary reverses comparisons (5 of its 124 changes), and the query can target the wrong person. SACC, a traceable editor we build as the instrument, lets us ablate two editing principles and show where its advantage does not hold.

Shuffled OPSD: When Semantic Consistency Hurts On-Policy Self-Distillation?

Qingyuan Zheng, Zhihao Dou — Wed, 04 Nov 2026 00:00:00 +0000

93% · New model 93%, Evaluation 40%

Knowledge distillation improves a student model's reasoning ability through supervision from a teacher. On-policy self-distillation (OPSD) uses the same model as both teacher and student under different contexts: the student generates reasoning trajectories from the question alone, while the teacher additionally receives a complete worked example and provides token-level supervision along those trajectories. Because the problem statement is shared, its content and ordering shape both student generation and teacher feedback. Preserving its semantic structure therefore seems natural, but does it always improve learning? We introduce Shuffled OPSD, which shuffles the words of the shared problem statement while keeping the teacher-only worked example intact. Across three mathematical reasoning benchmarks and three Qwen3 model scales, the method improves mean accuracy over both an intact-target control and Standard OPSD under matched training configurations. The largest gain over Standard OPSD is 3.52 percentage points on Qwen3-1.7B. Further controls show that preserving larger word blocks offers no advantage over word-level shuffling at the evaluated scales, whereas shuffling the teacher-only worked example reduces accuracy at every scale. Cross-problem mixing and target-removal experiments further distinguish the roles of problem content, word order, and teacher context. These results show that disrupting the shared problem statement can improve learning, while disrupting the teacher-only reference hurts performance. They challenge semantic coherence as a universally beneficial principle for OPSD and motivate closer study of how input structure shapes student generation and teacher supervision.

Axion 1.0: Open Agent Foundation Models, Simplified

Yuqi Jia, Yunjie Tian, Jiayu Zheng, Yijun Liang, Jingchen Sun, Mengyu Zhao — Wed, 04 Nov 2026 00:00:00 +0000

93% · New model 93%, Agents 93%

We introduce Axion 1.0, a fully open agent foundation model that combines strong and broad agentic capabilities with a remarkably simple training recipe. Axion 1.0 is trained in only three stages: two lightweight supervised fine-tuning stages followed by reinforcement learning, without dedicated mid-training, direct preference optimization (DPO), or on-policy distillation (OPD). The first SFT stage develops strong self-contained reasoning through model-aware data construction, where we synthesize challenging problems and adversarially strengthen them based on the base model's own successful solution strategies. The second stage extends these capabilities to tool-dependent reasoning through a scalable pipeline for constructing diverse and automatically verifiable tasks grounded in real-world data sources and deterministic computation. Finally, reinforcement learning further strengthens long-horizon reasoning and interaction with external tools and environments. Despite this compact training recipe, Axion 1.0 achieves strong performance across a broad suite of agentic benchmarks and is competitive with leading commercial models on several of them. Equally importantly, Axion 1.0 is open beyond model weights: we release the complete recipe, including data construction, generation, filtering, verification, prompts, and training code, enabling the entire development process to be reproduced and extended. Axion 1.0 demonstrates that strong agent foundation models can be both open and simple, offering a practical and reproducible path toward advancing open agentic intelligence.

Does the Trust Boundary Hold over Time? A Long-Horizon Benchmark for Receiver-Directed Federated Learning

Ting Xu, Henry Leung — Wed, 04 Nov 2026 00:00:00 +0000

92% · Evaluation 92%

In receiver-directed federated learning, each site draws a trust boundary. For every source and model component it judges, on its own data, whether the transfer helps (admit), harms (reject) or cannot yet be told (abstain). We ask whether this boundary holds over hundreds of rounds. Our benchmark has a federation of thirteen semi-synthetic sites, an exact oracle for every evaluated transfer, paired seeds, an attack protocol, fourteen methods over 600 rounds and Qwen3-1.7B LoRA sites over 500. The answer depends on which boundary is measured. Across 20 paired CPU seeds, a gate judging against a fixed reference better identified transfers useful relative to the site's starting model. Yet it admitted 29% of transfers harmful to the model the site now has. A gate judging against the current model admitted 2% of harmful transfers on that target in the CPU tier, but missed many transfers useful to the starting model. Two short propositions explain these distinct failures. We also evaluate a held out audit that leaves gate verdicts intact. After training, it decides whether each final model is deployed or the site falls back to local training, and reports admissions that have gone stale. With independent audit data its guarantees hold after any training history. At 600 rounds the audit deployed 65 to 72% of models, none with negative benefit on its target, and kept 98% of the gain. No stale report had the wrong sign in our runs. With 20,000 audit observations, a variance-adaptive bound found 99% of stale transfers that cost at least 0.01 in log loss. A common-target Qwen3 comparison on 60 selected transfers shows the same tension over 100 rounds. With 38% adversarial sites, FedFomo and CS-Gate were the only evaluated methods with positive mean gain in every honest group.

ValQID: Learning the Net Value of Semantic Query Groups for Web Device Identification

Wed, 04 Nov 2026 00:00:00 +0000

92% · Efficiency 92%

Web interfaces are widely used to access network services and manage networked devices. Identifying the vendors and types of devices behind Web interfaces is a critical step in asset inventory, security management, and risk assessment. Device identity cues are dispersed across page text, configuration fields, and management functions, and interpreting them often requires linking related Web objects. Responses and objects require different amounts of semantic analysis, so exhaustively processing all candidate content can introduce unnecessary computation. We present ValQID, which uses a relation graph to combine dispersed device identity cues, evaluates semantic questions efficiently with DSEJEV, and accounts for complementarity among semantic questions when estimating computation value. Based on the current identification state, ValQID weighs each question group's expected identification benefit against its execution cost and adaptively allocates computation to identify device vendors and types. Evaluation on real Web response data shows that ValQID achieves relative gains of up to 47.1% in joint macro-F1 and up to 82.1% in macro-F1 for low-frequency vendors over a state-of-the-art method. DSEJEV speeds up semantic processing by up to 8.58 times compared with general-purpose language models.

Best-of-Depth: Learning Exploratory Dynamics in Looped Transformers

Shota Takashiro, Takeshi Kojima, Masanori Koyama, Kohei Hayashi, Yusuke Iwasawa — Wed, 04 Nov 2026 00:00:00 +0000

92% · New model 92%, Evaluation 34%

Looped Transformers extend inference-time computation through repeated latent-state updates, but final-depth training does not directly encourage useful intermediate alternatives. We introduce Best-of-Depth (BoD), which trains a recurrent trajectory as a candidate set by minimizing prediction loss across depths. Different valid targets can be matched at different depths without architectural changes or depth-specific labels. An exact-match analysis connects this objective to solution coverage. On N-Queens completion and graph coloring, BoD improves coverage over final-depth and intermediate-supervision objectives. N-Queens coverage reaches 91.03%, versus 45.75% for final-depth training. Further experiments show discovery beyond the training depth and recovery of solutions withheld from teacher pools on familiar inputs. Fine-tuning a pretrained looped language model with BoD yields higher pooled code-generation success than final-depth training when programs are generated separately at evaluated fixed depths. These results support training recurrent computation for solution discovery along its trajectory.

ReSI-Math: Exploring Recoverable Capability Frontiers for Efficient Reasoning Improvement

Yao Cui, Zhen Yang, Yuandong Wang, Yangfu Zhu, Tianyu Wo, Zhenzhou Shao — Wed, 04 Nov 2026 00:00:00 +0000

92% · New model 92%, Evaluation 55%, Efficiency 30%

Reinforcement learning with verifiable rewards (RLVR) has become an effective approach for improving mathematical reasoning in large language models. However, a fundamental question remains: when a model fails to solve a problem, does it truly lack the underlying capability? In this work, we reveal that direct failure does not necessarily imply inability. Some reasoning failures can be recovered by exploring the model's own trajectories, exposing a recoverable capability frontier: a policy-dependent region where solutions are not reliably generated through direct decoding but can be elicited through trajectory-level search. Based on this observation, we introduce ReSI-Math, a framework that explores recoverable capability frontiers and converts recovered capabilities into data-efficient RLVR supervision. ReSI-Math uses Self-Judge-guided evolutionary search to identify recoverable failures under a fixed policy and adaptively selects frontier problems for sequential RLVR as the policy evolves. On Qwen3-VL-8B, correct-incorrect trajectory crossover produces 272 verifier-confirmed correct offspring from 621 attempts (43.8%). After three sequential one-shot RLVR stages, the MATH500 and WE-MATH average improves from 78.71 to 80.16. Under matched single-problem RLVR, random selection reaches only 53.31 versus 54.16 for frontier-guided Q1.

Suppressing Misaligned Behaviors in Language Models, Durably

Luxi He, Pengcheng Jiang, Jifan Zhang, Jiawei Han, Danqi Chen — Wed, 04 Nov 2026 00:00:00 +0000

92% · New model 92%, Evaluation 84%

Undesired model behaviors can emerge or re-emerge at any stage of training. Once observed, how effectively can we suppress them? To study this question, we first construct two model organisms--language models deliberately trained to exhibit a specific misalignment--that differ in what triggers the behavior. The narrow-trigger organism misbehaves only when the prompt contains a specific cue: it cheats on multiple-choice questions by following a planted hint that points to a wrong answer, and behaves normally otherwise. The broad-trigger organism has no such cue: after fine-tuning on insecure code, it gives harmful responses across a wide range of open-ended prompts. We then formalize post-hoc behavioral suppression and evaluate methods along three criteria: immediate suppression of the target behavior, durability under further training, and preservation of general capability. We compare standard methods--SFT, GRPO, and gradient-ascent unlearning--along these dimensions. Building on prior work on alignment shallowness, we hypothesize that more durable suppression requires the ability to recover from a wider range of misalignment states. We introduce Prefix-Expanded Adversarial Reinforcement Learning (PEARL), a GRPO variant that augments rollouts with continuations from intermediate states sampled along cached organism trajectories. Across both organism settings and two model families (Qwen3-4B, GPT-OSS-20B), PEARL achieves strong immediate suppression while preserving task accuracy, and remains durable under further fine-tuning on either the organism's data or unrelated data. It obtains the greatest benefit when the undesirable behavior is broad and cannot be addressed through targeted SFT. Together, our framework and method offer a step toward more durable post-hoc removal of undesired behaviors.

Training LLM Agents to Automate LLM Post-Training with RSI-Dojo

Shuzheng Gao, Zhaojian Yu, Penghao Yin, Yuling Shi, Kai Cai, Shilin He — Wed, 04 Nov 2026 00:00:00 +0000

91% · New model 91%, Evaluation 80%

Recursive Self-Improvement (RSI) aims to enable AI systems to autonomously build and improve AI systems themselves, including models, harnesses, and underlying infrastructure. Within this broader agenda, a central capability is enabling agents to train and improve base models. However, existing studies on automated model training primarily focused on workflow- or harness-level optimizations, rather than internalizing the ability to conduct post-training into the model itself. Developing this capability requires executable model training environments that can produce high-quality trajectories demonstrating how to select data, train models, diagnose failures, and iteratively improve results. In this paper, we introduce RSI-Dojo, a collection of executable model training environments designed to teach LLM agents to conduct post-training. To construct RSI-Dojo, we develop RSI-Craft, an automatic pipeline that converts human research artifacts into executable model training environments. RSI-Craft automatically collects research papers from top venues, instantiates datasets and metrics into a unified interface, and reproduces evaluation dependencies. It then filters tasks by data size, execution reliability, and discriminability, yielding RSI-Dojo with 281 seed model training tasks. Building on RSI-Dojo, we employ a trajectory collection method with an improved harness to collect high-quality model-training trajectories from frontier models, then retain examples through rubric-based and outcome-based filtering. We fine-tune Qwen3.5-9B and Qwen3.6-35B-A3B on these trajectories. The resulting RSI-Dojo-9B and RSI-Dojo-35B-A3B improve over their base models by 8.74 and 16.87 PostTrainBench points, respectively. RSI-Dojo-35B-A3B reaches 29.95 on PostTrainBench and 53.29 on PostTrainBench-Lite, matching the performance of frontier models such as GLM-5.2 and GPT 5.5.

Rethinking Next-Visit Diagnosis Prediction Beyond Aggregate Metrics: Retrieval--Memory Multi-Agent Adjudication

Heejoon Koo — Wed, 04 Nov 2026 00:00:00 +0000

91% · Evaluation 91%, Hallucination 36%

Next-visit diagnosis prediction forecasts the diagnoses documented at a patient's next encounter, turning retrospective electronic health records into prospective clinical prediction. However, conventional aggregate evaluation obscures its multifaceted nature: distributionally, models must handle variable-size diagnosis sets and long-tailed labels; temporally, they must distinguish persistent, newly recorded, and recurrent diagnoses; and both vary with observed history depth. We introduce an evaluation framework along these dimensions and apply it to rule-based heuristics, LLM-based prompting, retrieval, memory, and multi-agent approaches. This reveals that repeating the last visit outperforms most LLM-based approaches overall despite predicting no new or recurrent diagnoses, and that retrieval and memory exhibit distinct profiles whose relative strengths vary across temporal transitions and backbones. Motivated by this, we propose LONGINUS, a retrieval-memory multi-agent adjudication framework combining evidence retrieved from patient records with structured memory of the patient trajectory. Using cross-view agreement as a reliability heuristic, it preserves shared predictions and adjudicates disagreements without revealing their source. Across MIMIC-III and MIMIC-IV with four LLM backbones, LONGINUS achieves the best transition-specific performance among LLM-based methods in nearly all settings while remaining competitive with them on aggregate metrics.

RAPID: Increasing Efficiency of Internal Distillation via Randomized Methods

Ryan Swift, Konstantinos Psounis — Wed, 04 Nov 2026 00:00:00 +0000

91% · New model 91%, Efficiency 46%

Knowledge distillation is a technique used to transfer the expertise of large teacher models into small, efficient student models. Recent advances have integrated the distillation of internal activations, more directly training students to learn similar representations as the teacher. One of the core challenges of internal distillation is the dimensional incompatibility between the student and teacher internal representations. Existing methods add learnable matrices to project the larger teacher activations into the smaller representation space of the student. This is a notable point of inefficiency, as learning additional parameters for the projectors increases training costs, only for them to be discarded immediately after training. We introduce RAPID (RAndomized Projection for Internal Distillation) a novel distillation methodology which eliminates the wasteful projection learning by leveraging randomized methods. RAPID utilizes an efficient transform construction based on the Johnson-Lindenstrauss lemma to compress the teacher representations into the student embedding space with strong theoretical guarantees of structural preservation. RAPID also employs a combination of exponential moving average (EMA) and homoscedastic uncertainty weighting in a novel multi-objective function designed for internal distillation. Our experiments in post-training for reasoning demonstrate that RAPID increases training efficiency by as much as 21.1% while maintaining performance comparable to existing methods.

Computer-Use Agents Gain Little Benefit from a World Model When Conditioned on Tutorials

Anh Nguyen, Tianyi Xu, Phat Tran, Harry Yu, Huazheng Wang, Stefan Lee — Wed, 04 Nov 2026 00:00:00 +0000

90% · Evaluation 90%, Agents 83%

Recent work has improved computer-use agents (CUAs) by combining retrieved task-relevant tutorials and look-ahead planning using world models with MLLM judges. However, the contribution of each component -- tutorials, the judge, and the world model -- to overall improvement has not been evaluated in isolation. To fill this gap, we first construct OSWorld-Tutorial and ScienceBoard-Tutorial: two datasets for CUAs with curated multimodal tutorials, covering everyday and scientific domains. We then propose RAG-PolicyJudge (RAG-PJ), a policy-judge framework for CUAs conditioned on task-specific tutorials: the policy proposes candidate actions, and the judge selects among them. To isolate the effect of world models given tutorials, we compare RAG-PJ with and without common world models: using either zero-shot text-based descriptions of next states or an oracle that returns ground-truth visual observations. We find that RAG-PJ gains little benefit from these rollouts when tutorials are present, even with perfect visual futures. This suggests prior world-model gains in tutorial-conditioned methods come mainly from tutorial understanding and the judge's ability to select the proposed actions, rather than from predicted rollouts. Error analysis reveals that the bottleneck is not rollout fidelity, but the judge's ability to reason over future states and recognize task completion. This highlights the need to improve not just next-state prediction in world models, but also reasoning over their predicted rollouts. Datasets and code are available at https://github.com/realjoenguyen/RAGPJ-official.

Unifying and Beyond the Theory of Pervasive Capability Degradation in LLM Post-training SFT from Feature Interference

Hang Chen, Jiaying Zhu, Wenya Wang — Wed, 04 Nov 2026 00:00:00 +0000

90% · New model 90%, Evaluation 40%

Pervasive Capability Degradation (PCD) refers to the phenomenon in which supervised fine-tuning (SFT), while adapting a large language model (LLM) to a target task, often erodes the general capabilities acquired during pre-training. Prevailing explanations attribute PCD to macroscopic distribution shift and therefore lack microscopic manipulability, leaving existing mitigation methods without a complete theoretical foundation. In this paper, we first develop a bottom-up geometric theory of PCD grounded in feature interference under superposition: narrow-domain SFT data disrupt the almost-orthogonal feature structure formed during pre-training, and the resulting interference drives activations across decision hyperplanes in the output Laguerre-Voronoi geometry, flipping predictions on general tasks. This theory unifies existing methods into six paradigms and reveals both the source of their efficacy and their theoretical limits. We then prove that cross-entropy under one-hot supervision cannot resolve PCD on its own, and accordingly propose the AEGIS loss, which directly constrains the drift of activation margins along all competing token directions, thereby decoupling interference suppression from task adaptation. Across 3 model families and 14 SFT tasks, AEGIS outperforms representative methods of every paradigm on target tasks and pervasive capabilities simultaneously, at only about 1.3 times the computational cost of standard SFT. Its generality, scalability, and mechanistic interpretability offer a new perspective on fully resolving PCD.

Route Anything: A Study of Model and Policy Routing for General-Purpose AI Systems

Pengfei Zhou, Yixing Ma, Hexin Wang, Weidong Tang, Zhenglin Wan, Wangbo Zhao — Wed, 04 Nov 2026 00:00:00 +0000

90% · Evaluation 90%, Efficiency 87%

Artificial general intelligence (AGI) requires a method that can understand, generate, reason, and act across diverse domains and modalities. Rather than relying on a single model to master every capability, a promising path toward AGI now is to orchestrate a heterogeneous ecosystem of specialized models, tools, and policies (candidates) into a unified AI system. At the heart of this vision lies a fundamental question: how can an AI system dynamically select and coordinate the most capable candidate for each task while balancing performance and computational cost? In this work, we present Route Anything, a unified framework for studying model and policy routing across general-purpose AI systems. We develop a unified theoretical foundation that characterizes optimal routing under quality--cost trade-offs and establishes how the information available to a router determines its attainable performance. To systematically investigate routing across diverse capabilities, we first construct an omni-modal trajectory matrix spanning 25 benchmarks across language understanding, image, video, and speech generation, coding and workplace agents, and robotic manipulation, comprising 374,401 scored runs from 452 candidate configurations. Building on this framework, we train and test different routers to investigate the potential of omni-modal model routing, the capabilities of learned routers, and the conditions under which routing improves efficiency. Our results reveal substantial opportunities for capability orchestration: per-input oracle routing outperforms the best fixed candidate by 5.4 points, and learned routers outperform it by 0.3 points across 16 benchmarks, by 2.1 points on video generation, and by 1.1 points on robot manipulation while using 24% less policy time. More importantly, routing raises efficiency by 8.82x on average compared with the reference model.

★ NorthCUA Technical Report: Turbocharging Computer-Use Agents through Experience at Scale

Chaoyun Zhang, Bingxuan Li, Pinxue Guo, Yangfan Gao, Zikun Cai, Tianci Xie — Tue, 03 Nov 2026 00:00:00 +0000

95% · Agents 95%, New model 90%, Efficiency 51%

Scaling computer-use agents from individual actions to complex software work requires experience that connects decisions to complete outcomes. We present NorthCUA, an integrated system for turbocharging computer-use agents through experience at scale. Its Experience Flywheel supplies diverse executable tasks to the NorthCUA Harness, which sustains long interactions, records decisions, and checks task outcomes. The Flywheel uses these records to curate verified trajectories. For unsolved tasks, recursive self-improvement evolves reusable skills and preserves verified milestones so later attempts build on earlier progress. Staged Learning combines this experience with grounding data from the Flywheel to develop GUI competence through mid-training and long-horizon supervised fine-tuning. The policy then learns from fresh attempts in the same Harness through joint desktop-web reinforcement learning, closing the loop between experience production and policy improvement. The resulting Nano, Mini, and Pro family achieves state-of-the-art results among open-source models across GUI grounding, desktop execution, and Web interaction under the reported evaluation settings. For more responsive Web workflows, NorthCUA-Ultrafast pairs direct action selection with open-ended generation on pre-RL Nano, offering a measured quality-latency tradeoff. Together, these advances establish NorthCUA as a high-performing model family and a reusable foundation for agents that carry complex software work from intent to completion.

★ Apriel Voice 27B Preview: An Open Model for Speech, Reasoning, and Tool Use

Tue, 03 Nov 2026 00:00:00 +0000

95% · New model 95%, Agents 68%, Evaluation 57%

A capable voice assistant must do more than recognize words; it must understand requests, reason over context, sustain conversations, and act through tools. We present Apriel Voice 27B Preview, an open speech-language model that unifies these capabilities within a shared reasoning backbone while retaining text and image input modalities. Built on the post-trained Gemma 4 26B-A4B checkpoint, Apriel Voice directly processes speech and generates both responses and tool calls without requiring a separate dialogue model. We introduce a two-stage training procedure to adapt the model to speech. The first stage establishes the speech interface, while the second jointly adapts the projector and backbone through supervised fine-tuning, text replay, and source-dependent cross-entropy and forward Kullback--Leibler divergence supervision. Automatically constructed speech-conditioned dialogues provide conversational training data, while local teacher serving, overlapping teacher and student computation, and fused full-vocabulary losses enable efficient online distillation. In thinking mode, Apriel Voice achieves a score of 98.90 on Big Bench Audio, 81.13 on Speech-BFCL, and a task success rate of 65.51% on τ-Voice, with a mean latency of 1.91 seconds. It also achieves 51.27% English EVA accuracy and 74.13% experience pass@1. Finally, we deploy Apriel Voice 27B Preview on a custom serving platform to support interactive voice interactions.

WeFair: Fidelity-Aware Alignment from Human Feedback for Personalized Query Rewriting in Short-Video Search

Yingtao Ren, Aolin Li, Cheng Cheng, Huiyun Hu, Huasheng Liang, Baoxing Xie — Tue, 03 Nov 2026 00:00:00 +0000

93% · New model 93%, Evaluation 43%

Short-video search needs compact query rewriters that recover users' personalized intent from ambiguous queries under strict latency constraints. Existing LLM-based methods have improved rewriting by using synthetic reasoning and subsequent consumption-behavioral feedback. However, overemphasizing subsequent consumption can favor engaging content over the user's immediate need, causing intent drift even when rewritten queries are grounded in behavioral history. We propose WeFair, Fidelity-Aware Alignment from Human Feedback for Personalized Query Rewriting in WeChat short-video search. WeFair uses human feedback according to how directly it reflects personalized intent, assigning feedback fidelity to a progression from demonstration to preference to verification. We first mine actual queries and user reformulations from search sessions, and then screen them for contextual support. In the demonstration stage, a supervised fine-tuning curriculum progresses from large-scale context-to-query learning to reasoning and direct prediction on human reformulation and preservation. In the preference stage, direct preference optimization learns relative preferences from subsequent consumption feedback. In the verification stage, group relative policy optimization rewards exact agreement with reformulation or preservation targets to constrain intent drift. Further analysis shows that broad supervision benefits the full curriculum supervised training, while strict verification achieves better performance than auxiliary rewards. In offline experiments, WeFair-2B outperforms strong rewriting baselines in overall accuracy and intent preservation. An online A/B test shows relative increases of 2.593% in effective video-view rate and a 2.350% relative reduction in query reformulation rate. WeFair-2B has been deployed in the WeChat short-video search system, which serves over 100 million queries per day.

CUA-Master: Scaling Computer-use Agents through Open-World Recursive Improvements

Xinlei Yu, Gen Li, Ji Zhao, Zhangquan Chen, Congcong Wang, Shuai Dong — Tue, 03 Nov 2026 00:00:00 +0000

91% · New model 91%, Agents 90%, Evaluation 44%

Despite recent advances in CUAs, their development remains constrained by the limited diversity, controllability, and reliability of interaction data. Moreover, conventional optimization pipelines lack iterative feedback, while existing recursive improvement approaches often rely on self-generated experiences, limiting ability expansion and yielding diminishing returns. To address these challenges, we introduce CUA-Master, a nested double-loop recursive improvement framework. At the system level, an outer world-data-model flywheel continuously incorporates open-world corpus and leverages evolving model weaknesses to guide subsequent data construction and optimization iterations. At the data level, an inner flywheel couples open-world seeding and autocatalytic propagation. The former mines authentic workflows as seed tasks from instructional videos, and employs visual-semantic dual-path trajectory reconstruction; The latter transforms these seeds into reusable primitives for controllable synthesis of diverse, complex, and verifiable tasks. To enable efficient and reliable trajectory collection, we further introduce a feedback-guided snapshot-rollback retry mechanism in various real environments. Extensive experiments demonstrate that CUA-Master-27B achieves competitive performance across GUI-only and GUI-CLI hybrid benchmarks, scoring 80.2 on OSWorld Verified, 36.9 on OSWorld 2.0, 43.9 on WeaveBench, and 25.0 on Agents' Last Exam. Our framework also substantially improves the quality and efficiency of data synthesis, and exhibits continuous gains across recursive iterations.

★ P-ContextBench: From Personal Fact Recall to Personalized Agent Behavior

Grant Nasir Brown, Abdul Basir Bari, Anmol Tadikonda, Isabel Song — Mon, 02 Nov 2026 00:00:00 +0000

99% · Evaluation 99%, Hallucination 52%

Large language models (LLMs) are increasingly used for personal tasks, yet evaluating their safe use of personal context remains difficult because such contexts are private, fragmented, dynamic, and socially situated. We introduce P-ContextBench, a realistic, verifiable, and scalable benchmark for evaluating personal-context understanding. P-ContextBench constructs population-structured, life-story-grounded digital humans from demographic seeds reflecting the U.S. population. Rather than using explicit, static profiles, P-ContextBench embeds traits, routines, goals, and constraints as implicit evidence across fragmented, evolving synthetic mobile-app records. It evaluates five capabilities: single-source fact recovery, multi-source fact recovery, temporal state tracking, inference of higher-level personal patterns from routine behavior, and safe handling of requests conflicting with persona-specific constraints through compliance, warning, clarification, or refusal. A multi-stage validation pipeline checks persona plausibility, context recoverability, information leakage, evidence preservation, test-case grounding, and human-judge agreement. We instantiate P-ContextBench with 100 personas, 10,088 hidden facts, 18,261 app-trace fragments, and 4,810 verified test cases. The pipeline supports configurable generation across scales and demographic groups. Cost-effective state-of-the-art (SOTA) LLMs relevant to personal agents achieve below 60% accuracy on temporal tracking, higher-level pattern inference, and constraint-aware action-boundary decisions, with fewer than 10% of test cases answered correctly across all three repeated runs. Even a high-performing SOTA model remains inconsistent on these capabilities. P-ContextBench provides measurable targets for improving the accuracy, reliability, and safety of LLM-based personal agents.

★ Do Open Language Models Invent Scripture? BILAL-BENCH: Benchmarking Islamic-Source Fabrication in Bengali

Sabah Nushra, Md Kishor Morol — Mon, 02 Nov 2026 00:00:00 +0000

99% · Evaluation 99%, Hallucination 97%

When a language model invents a Qur'anic verse or hadith, it gives false authority to text that readers may treat as sacred. Existing Islamic hallucination benchmarks are largely Arabic and often closed-ended, while Bengali resources mainly study general-domain hallucination or factual knowledge. We introduce BILAL-BENCH, a 240-question expert-verified benchmark for open-ended Islamic-source fabrication in Bengali. It covers Qur'anic attribution, hadith, fiqh, contemporary issues, and adversarial false-premise prompts, and uses a two-axis rubric that separates source errors from assertiveness. We evaluate seven open-weight models from 1B to 120B on 1,680 double-annotated responses. Six of the seven models confidently fabricate on 79.3-95.3% of source-production questions, with no monotonic improvement from scale. Hadith provenance is the weakest category across models: only two of 315 responses are correct and correctly sourced. The two 8B BanglaLLaMA variants have similar confident-fabrication rates but different error profiles. Models vary far more in whether they withhold a claim they cannot support than in source accuracy: appropriate abstention or false-premise correction ranges from 6 to 46 responses across the benchmark, and does not track parameter count.

★ Do Language Models Know What Compelling Scientific Evidence Looks Like?

Conor Igoe, Gabriel Grand, Zane Koch, Albert Bou, Eli Moss, Michaela M. Hinks — Mon, 02 Nov 2026 00:00:00 +0000

99% · Evaluation 99%, New model 71%, Efficiency 41%

Expert scientists excel at weighing empirical findings from multiple sources that vary in both statistical reliability and epistemic relevance. To what extent are language models, which are increasingly driving autonomous discovery systems, capable of such evidence deliberation? We introduce EDB-Bio, a scientific reasoning benchmark curated from real-world data in three biological domains: drug-perturbation transcriptomics, CRISPR screens, and human genome-wide association studies (GWAS). Each of EDB-Bio's 96k episodes presents a sequence of experimental results from published studies and asks the model to predict the probability that a target hypothesis is true. Out-of-the-box, open-weight and frontier models alike demonstrate systematic underconfidence and attain Brier scores comparable to a trivial base-rate forecaster. Post-training against strictly proper scoring rules significantly improves both calibration and discrimination. This calibration pays off downstream: in a sequential experiment selection task, our Ibis0 model is 2.5 times more efficient than the best frontier model. Throughout, quantifying evidence weights as Bayes factors reveals how models update their beliefs over time, whether they agree with expert biologists on relevance, and where they exhibit inconsistencies and biases. Our work extends calibration training and uncertainty analysis to a realistic scientific forecasting setting with applications to human cancer biology. We release all code, data, and trained models.

★ ReplayRCA: Measuring Post-hoc Diagnosis and Live Oversight of Agent Failures by Counterfactual Replay

Jeremy Canale, Fabien Soulis — Mon, 02 Nov 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 58%

When a language-model agent fails a multi-step task, can another model find the decisive mistake and fix it with a single action? We introduce ReplayRCA, a benchmark that scores such diagnoses by intervention rather than by agreement with annotations: the environment is restored to the state preceding the chosen step, the proposed action replaces the agent's own, and the same agent continues. The score, causal repair, is the continuation success rate minus that of a paired placebo that replays the agent's original action through the same mechanism. Two tracks share this protocol: a post-hoc autopsy, where the diagnoser sees the whole failed run and is told that it failed, and live oversight, where a supervisor watches the run unfold and may intervene once, before an irreversible action, without knowing whether the run will fail. Tasks come from OpsWorld, a deterministic, procedurally generated back-office environment with six difficulty levels, the last of which is multimodal (QR codes, handwriting). On 62 calibrated systematic failures, the best of 13 frontier models (Grok 4.7) reaches a causal repair of +51.3 points; the top eight models are statistically indistinguishable, the flagship models of the two largest vendors rank 8th and 12th, and 5 items are repaired by no model. Forcing a strong model to intervene at the last step costs it 17.1 points, so the last-step shortcut of annotation-based attribution benchmarks does not transfer. On 87 runs, the two best live supervisors are statistically tied and one of them is a small open-weight model (DeepSeek V4.1 Flash); the supervisor that is among the best post-hoc diagnosers (GPT-6.1 Sol) is the weakest live one (-14.5 points on the same runs, p = 0.002), so post-hoc diagnostic performance can be a poor proxy for live oversight. The environment and the evaluation code are released at https://github.com/jeremy1392/RCAreplay.

★ BABA-Bench: Evaluating Long-Horizon Patient-Advocacy Agents

Shriram Vasudevan, Videet Mehta, Connor Sweeney — Mon, 02 Nov 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 55%

An LLM agent that helps a patient get care must do two things. It must reach the correct outcome, and it must follow the necessary procedure. For example, the agent must get the patient's permission before it shares the patient's records. We introduce BABA-Bench, a benchmark of 47 interactive worlds. We built each world from a de-identified, real patient advocacy case that expert advocates reviewed. In each world, the agent must find hidden information and coordinate insurers, clinics, suppliers, pharmacies, and community organizations. The engine checks each action at the time the agent does it. It checks if the agent has permission, if the agent has evidence for the action, and if the action is prohibited. We evaluated 14 frontier models with three runs on each world, for a total of 1,974 runs. Agents often reached the correct outcome by an invalid route. Of the 921 runs that reached all required outcomes, 254 (28%) did not follow the necessary procedure. Claude Opus 5 reached all required outcomes in 81.6% of its runs, but only 52.5% of its runs passed all checks. When we gave two models a verified route, their pass rates increased from 29.8% to 91.5% and from 44.0% to 97.2%. For these models, the main problem is to find a route, not to execute it.

★ FrontierMat: Benchmarking Language Models at the Frontier of Materials R&D

Feng Chen, Jianguo Huang, Yushan Xiao, Yue Li, Bijun Tang, Xuyu Dong — Mon, 02 Nov 2026 00:00:00 +0000

99% · Evaluation 99%, Hallucination 65%, New model 54%

To accelerate materials discovery, language models must be able to address problems at the research frontier. Yet textbook- and examination-derived benchmarks primarily assess established knowledge, leaving performance on frontier research problems largely untested. We introduce FrontierMat, a benchmark of 330 problems drawn from frontier academic and industrial research in fields including aerospace, semiconductors, embodied intelligence and nuclear fusion. Every problem is validated by leading experimentalists and industry experts in its field. Each problem presents a real study and asks the model to infer a withheld outcome or draw a conclusion from selected experimental evidence. Problems are annotated by functional domain, material system and the scale of the governing mechanism; 130 require cross-scale reasoning. The strongest general-purpose model scores only 53.7%, and on matched problems, cross-scale problems score 8.7 points below single-scale ones. Half of the problems are solved by at most one model, and models fail chiefly by endorsing what is plausible rather than what was observed. A domain-adaptation study with meta-weighted continued pretraining and scale-modular self-distillation raises an open-weight model from 36.7% to 55.4%, matching the strongest proprietary model, with improvements on public materials benchmarks and no decline in aggregate performance on general benchmarks. FrontierMat provides a testbed for measuring progress toward language models that support frontier materials R&D.

★ GraphNIAH: A Benchmark for Long-Context Reasoning over Real-World Graphs

Minh-Nam Tran, Abulhair Saparov — Mon, 02 Nov 2026 00:00:00 +0000

99% · Evaluation 99%

The needle-in-a-haystack (NIAH) benchmark was introduced as a test for measuring a model's ability to find small relevant facts within a very large body of input text. However, existing NIAH benchmarks each suffer from one or more of the following limitations: (1) reliance on synthetic data, so their evaluations may not generalize to real-world settings; (2) reuse of previously published datasets, leading to the risk of benchmark contamination; (3) dependence on manual annotation, which requires human effort to create and update data and is prone to future benchmark contamination; and (4) unrealistic simplification of long-context multi-hop reasoning where each next step is trivial given previous steps, while real-world reasoning often requires planning over multiple possible proof paths. To address these limitations, we introduce \Benchmark, a dynamic benchmark built from real citation relationships among scientific papers published after the evaluated models' knowledge cutoff date, which utilizes real-world information while ensuring no benchmark contamination. The fully automated benchmark construction pipeline samples representative subgraphs of the citation graph formed by these papers and converts each subgraph into a long-context QA example with a verifiable answer. We use GraphNIAH to evaluate eight LLMs from the GPT-5.6, Gemini-3, and Qwen3.5 families. We find that accuracy falls as questions require longer reasoning paths: at question depths of 16 and 32, the GPT-5.6 and Qwen3.5 models answer YES/NO questions no better than random guessing, mostly by answering NO; four of the eight models answer no counting example correctly, and the others model get worse as the subgraph size increases. These findings suggest that current LLMs often stop searching the citation relationships in their context before they reach the answer, even though every fact they need is in the context.

★ PersonaChatBench: Evaluating Chat Assistants with Diverse Simulated Users

Zheyuan Deng, Taoran Li, Jingwen Zhang, Yuante Li, Dachuan Zhao, Fulei Zhang — Mon, 02 Nov 2026 00:00:00 +0000

99% · Evaluation 99%

Conversational assistants are expected to help people with diverse backgrounds, preferences, and communication styles. Yet evaluations centered on a single cooperative user provide limited insight into how reliably an assistant serves users with different personas. We introduce PersonaChatBench, a framework for evaluating task competence and responsiveness to diverse user needs. PersonaChatBench pairs the same underlying task with multiple simulated users, each guided by a private persona card that specifies distinct needs, preferences, and interaction behaviors. By holding the task fixed while varying the user, the framework examines how assistant performance changes across user conditions. We evaluate 20 language models across six domains, separately assessing native task outcomes, fulfillment of expressed persona needs, and simulator fidelity. Our results show that task performance and need fulfillment vary across personas and domains. Strong task performance does not necessarily translate into satisfying individual needs, and models' relative strengths in meeting those needs are not consistent across domains. PersonaChatBench makes these differences measurable, revealing gaps obscured by aggregate scores and helping identify which users an assistant serves well and where adaptation is needed. It provides a basis for evaluating consistency across users and guiding the development of assistants that better accommodate diverse needs in everyday applications.

★ More Rules, Longer Lists? Benchmarking Self-Evolving Agents on a Hidden-Rule Curriculum

Chenjie Qiu, Boyan Xu, Shaobin Shi, Bingfeng Chen, Ruichu Cai, Dongsheng Luo — Mon, 02 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

Self-evolving agents rewrite a persistent state from feedback, but it is largely untested whether this lets them learn a rule nobody told them. An agent can also pass by writing down each correction as an exception, and its list then grows with every new rule. We introduce CLR-Bench, a Curriculum of Latent Rules benchmark. Each chain adds one hidden rule per level on top of the earlier ones. A rule acts on an opaque label that stands for a named grouping published on the Web, such as the founding members of an airline alliance. The task gives one example value, practice feedback reveals one more member and one non-member, and the test checks unseen members and similar non-members. Probes certify that the base model knows the members of each grouping and does not guess it without feedback, and that the evidence identifies the grouping once labels are read as named groupings. An automated pipeline builds chains from a new data domain, and we release 20 chains over 10 domains. We evaluate six recent self-evolution methods. With the certifying model they barely learn. The best gain at most 5 points over a frozen agent and exclude only 6-12% of the unseen members from the count, fewer than the 20% reached by simply keeping the raw practice history. A fixed lesson that states the named-grouping convention lifts the same methods by 14.5 to 16.5 points. Their states hold the revealed member in 75-98% of cases and almost never the class name, and when practice exposes a member the list misses, they add that member to the list. With the stronger GPT-6-Sol, ACE and Dynamic Cheatsheet learn from feedback alone. They reach 78% accuracy and solve 59-60% of the single cells, against 53% and 8% with the certifying model, and they surpass the raw history. CLR-Bench therefore measures how well an agent learns, and this ability differs across models.

★ Beyond Plausible Prediction: Safety Sufficiency in Web World Models

Zhifei Hu, Alexandra I. Cristea, Adel Bibi, Jialin Yu — Mon, 02 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 86%, New model 41%

Web world models increasingly predict the consequences of candidate actions before agents act, yet a plausible predicted future does not necessarily support the correct safety decision. In a matched human evaluation of 3,000 predictions from ten frozen open models, 86.1% are judged semantically plausible. However, several models exhibit highly asymmetric authorisation behaviour, combining low false-safe execution with poor recall of clarifiable ASK cases and a strong preference for BLOCK. This motivates safety sufficiency: whether a predictive interface preserves and exposes the policy-decisive information required to distinguish actions that may be EXECUTED, require clarification through ASK, or should be BLOCKED. We introduce SafeWebWorld, a benchmark for evaluating safety sufficiency across five web-safety domains through action-conditioned futures, policy-decisive evidence, and policy-conditioned decisions. Under a controlled Security-Aligned Supervised Fine-Tuning protocol, evidence preservation can improve substantially while authorisation-sensitive decisions deteriorate. To determine whether this deterioration reflects loss of authorisation-sensitive score structure or a failure of the decision readout, we introduce Latent Authorisation Recovery (LAR). Frozen-score recalibration reveals substantial but backbone-dependent recovery, while paired recovery provides only a modest additional gain. Yet within-family and matched-authorisation evaluations remain poor, showing that aggregate recoverability does not imply robust fine-grained authorisation behaviour. Overall, plausible prediction, evidence preservation, and aggregate recovery are each insufficient on their own to establish robust authorisation.

★ Qistas: A Pre-Registered, Portable Protocol for Fairness Evaluation of Large Language Models

Nawar S. Alseelawi, Mustafa S. Aljumaily — Mon, 02 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

Fairness benchmarks for large language models (LLMs) are overwhelmingly English-born and American in their social categories; translated versions carry those categories into societies where they do not apply, and almost none fix their hypotheses before the models are run. We present Qistas (Arabic qistas, "the just balance"), a protocol for evaluating how LLMs treat the social, confessional, and regional constituencies of a society, portable across countries and instantiated first for Iraq. In its pilot form it measures only what needs no verdict on contested truth: procedural symmetry across matched counterfactual pairs, differential refusal, stereotype reliance under ambiguity, and terminological framing of documented events. Every measurement is paired by construction and analysed with paired machinery (sign-flip permutation tests, exact McNemar tests, template-clustered bootstrap intervals, Holm-corrected families), with a power analysis that states, before any data exist, the smallest effect each axis can detect after the design effect of template clustering. We release the authoring apparatus (a pruned 3x5 taxonomy with four cells closed by design, per-type checklists, a cumulative rulings log, a structural audit, and a human review notebook) and report its first field test: a semantic review of a 58-item seed bank in which 24 of 25 template groups failed at least one check. The defect patterns it exposed (evaluation criteria leaking into prompts, token-level swaps that break Arabic agreement, epistemically unbalanced pairs, single-theme refusal probes, unwritten dialect twins counted as coverage) are, we argue, general hazards of counterfactual fairness benchmarks. This is a protocol paper, not a results paper: no model has been evaluated, and the accompanying pre-registration commits the hypotheses, metrics, exclusion rules, and analysis plan that the pilot will follow.

★ Adaptive Personalization Under Preference Epistemic Uncertainty

Priyanka Dey, Rajarshi Das, Vikas Bahirwani, Qi Zeng, Julia Kruk, Qi Qi — Mon, 02 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, New model 57%

Human preferences change over time: an established preference may remain valid over time, become uncertain as context shifts, or be explicitly revised. Effective personalization requires conversational AI to adapt its response strategy to evolving preferences rather than relying on passive memory retrieval. We formalize this capability as adaptive personalization under epistemic uncertainty. A human preference study further confirms that people favor state-contingent response strategies as preference uncertainty changes. To evaluate whether language models mirror this adaptation, we create a multilingual benchmark spanning 250 personas, 5 languages, and 25K conversational turns that decouples strategy selection from response execution. Across eight closed- and open-weight models, we uncover a recognition-generation gap: models reliably identify appropriate strategies when selecting among candidate options, but collapse during open-ended generation, suffering a 35-46% drop in strategy calibration. Crucially, this failure is systematic: models default to exploit-first behavior, over-relying on existing preferences regardless of preference epistemic uncertainty. We show this failure is addressable: while prompting interventions fail, targeted training that supervises models' internal strategy representations toward state-conditional human preferences narrows the user acceptability generalization gap by a third on out-of-distribution personas and conversational styles.

★ Benchmarking Forecasting Agents on Yesterday's Tomorrows

Yuxuan Wang, Yishu Wang, Jiaqi Deng — Mon, 02 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 96%

Digital agents can now gather and organize information, reason over it, and use tools to complete complex tasks. Forecasting puts all of these abilities to the test: it requires integrating incomplete evidence, inferring causal drivers, and committing to calibrated probabilistic judgments. Existing benchmarks are poorly suited to measuring and improving this ability. Live benchmarks provide feedback only after events resolve, so every evaluation cycle waits on the calendar; fixed-corpus evaluations return results immediately but confine research to pre-collected documents. We introduce FABench, a scalable backtesting environment for Forecasting Agents. An automated pipeline discovers events, generates questions, and resolves outcomes for any past forecast date, so the benchmark can be renewed to stay ahead of model training cutoffs; its first release poses 287 questions across nine domains. Agents research each question freely in an archive environment that combines an offline archive with live web search screened to prevent temporal leakage. Evaluating agents built on eight recent models, we find that retrieval improves the Brier score of all eight in every retrieval setting and lifts the leading agents to the market consensus on the 54 questions linked to Polymarket, with the best three scoring slightly above it. Research traces reveal recurring failures in how agents weigh and apply evidence, leaving ample room for improvement. We release our code, data, and archive environment as a foundation for community research on forecasting agents.

★ Confident Fools or Careful Searchers? A Behavioural Study of When LLMs Search Beyond Their Static Internal Memory

Nipun D. Pathirage, Kwan Srijomkwan — Mon, 02 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 96%, Hallucination 51%

Agentic LLM assistants must repeatedly decide whether to answer from memory or to call a tool such as web search. Unnecessary searches cost latency and money; missing ones yield confident but unsupported answers. Prior work mostly measures verbalised confidence or trains new adaptive-retrieval methods, and rarely audits the search decisions that deployed products actually make. We introduce a benchmark of 400 multiple-choice questions that combines widely known inside-knowledge (IK) questions, which should be answered directly, with outside-knowledge (OK) questions about facts from news published after the models' training cut-offs (to our knowledge), which should trigger search or, without tools, abstention. Each item offers an abstain option, and two action-level scores, NAS and KIS, reward searching on OK questions and restraint on IK questions. Across five commercial LLMs with and without their vendors' web search, GPT-5 mini and Grok 4 Fast search on over 90% of OK but only 1% of IK questions, whereas Mistral Small searches on just 20% of OK questions, often abstaining instead of using its tool; one product's search indicator proved unusable. Searching is partly triggered by explicit temporal cues: on uncued OK questions, the NAS of the two selective models falls from about 0.72 to 0.57--0.58. The scores thus measure the search triggering of specific products, not metacognition in a strict sense.

★ ImplicitHop: Evaluating knowledge integration across context and memory

Alexander Detkov, Dashiel Karish, Matthew Oleynikov, Matt Thomson — Mon, 02 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

The ability to integrate new information and deduce its consequences is central to learning, understanding, and reasoning in large language models (LLMs). Whether information arrives during training or in-context, models must not only store local facts, but propagate their long-range implications globally. To explore information integration in LLMs, we introduce two complementary natural language benchmarks for large-scale evaluation of deductive reasoning as a function of inference depth. The first is a formal graph benchmark enabling controlled evaluation over a wide range of graph topologies and integration tasks; the second is a knowledge composition benchmark grounded in real-world knowledge graphs. We use these benchmarks to compare knowledge integration from information encoded in context versus information encoded in model weights during pretraining, evaluating a set of open models implicitly, without chain of thought. We find that accuracy decays sharply with increased composition length for both pretrained and in-context knowledge. Information presented in-context, however, substantially outperforms deduction over information encoded in weights. These results quantify a fundamental limitation of current models with direct implications for prompt-based control of agent behavior and for continual learning strategies, and they provide a foundation for developing training, prompting, and integration methods that support deeper reasoning.

★ TRIDENT and SHAKTI: A Framework and Benchmark for Hindi AI-Text Detection

Shifali Agrahari, Sujit Kumar, Sanasam Ranbir Singh — Mon, 02 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

As Large Language Models (LLMs) become increasingly multilingual, distinguishing human-written from AI-generated text is essential for maintaining information integrity. However, research remains English-centric, with limited benchmarks for low-resource languages such as Hindi. To address this gap, we introduce SHAKTI, a large-scale Hindi benchmark covering four genres, eight LLM families, multiple prompting strategies, and adversarial transformations. We evaluate its quality through human annotation and analyze linguistic, grammatical, and sentiment-based differences between human- and AI-generated texts. Building on these findings, we propose TRIDENT, a round-trip translation-based detection framework with three variants: TRIDENT-S (static threshold), TRIDENT-D (dynamic threshold), and TRIDENT-C (logistic regression using multi-distance RTT features). TRIDENT leverages semantic consistency between original and round-trip translated texts to identify AI-generated content. Extensive experiments show that TRIDENT outperforms conventional and supervised baselines, with TRIDENT-S achieving the best overall performance. Further evaluations across seven languages, adversarial settings, and unseen LLMs demonstrate its multilingual applicability and robustness, while highlighting sensitivity to the translation model.

★ SERIAL-7: Deriving, Realizing, and Preserving Machine-Verifiable Outputs Across Representations

Youngkyung Seo, MiHyeon Kim, Jisoo Lee, Hwijung Ryu — Mon, 02 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 41%

Large language models increasingly generate structured outputs consumed by programs and downstream systems, where semantic correctness alone does not guarantee reliable execution. We introduce SERIAL-7, a Korean-English benchmark built around an executable verification framework for machine-verifiable generation across seven representations. Using 177 validated canonical artifacts, SERIAL-7 separates semantic derivation (Derive), constrained artifact generation (Realize), and cross-representation preservation (Preserve). Its 43 atomic representation constraints are paired with executable checks over format, structure, and task-specific semantics, enabling progressively stricter verification from representation compliance to full artifact validity. Across ten models, REALIZE-L3 HSR ranges from 42.0% to 93.9%, while Full Compliance exposes additional artifact-level failures beyond correct final answers and valid representations. These results show that reliable structured generation requires explicit, executable verification of how artifacts are realized and preserved across representations.

★ From the Zoo to the BULLPEN: Sample-Efficient LLM Profiling via Multi-Channel Model Embeddings

Russell Ro, Qiming Cui, Reza Abbasi-Asl — Mon, 02 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Efficiency 83%, New model 49%

Choosing among the hundreds of thousands of available large language models (LLMs) requires a thorough understanding of each candidate model's properties. In practice, each property is measured by its own benchmark suite, so profiling a model on many properties means paying for many separate evaluations. To address this, model embeddings compress each model into a fixed-length vector from which many properties can be predicted. Existing methods, such as EmbedLLM, build these vectors from correctness patterns and question features, and often need large probe batteries to do so. However, many behaviors that are important when choosing a model, such as its verbosity, safety and style, show up more in how it answers than in whether it is correct. We introduce BULLPEN (Behavioral Understanding of Large Language models via Profile ENcodings), a framework that extends EmbedLLM in three ways. First, BULLPEN organizes the downstream tasks that prior work uses and adds 20 task benchmarks we built on behavior, style and provenance. Second, it is the first framework to read a model's own answer text jointly with its correctness and the question text. Third, it encodes models from a few hundred questions instead of the full 24,118-question pool. Across 260 open-weight models, our answer-text encoder scores 0.48 +/- 0.10 on the new tasks against 0.35 +/- 0.10 for the best prior encoder. Our encoders that read answer text keep at least 91% of their full-pool skill from only 400 questions (1.3% of the inference compute of the full question pool), while a correctness-only baseline keeps 83% with the same 400 questions. Because these embeddings need only model outputs, they support behavioral auditing and low-cost profiling, and could extend to API-only models. Code and data: https://github.com/wcquesera/bullpen

★ BlindToChange: AI Agents Can Find Changes When Asked But Do They Look When Needed?

Anirudh Phukan, Tanya Goyal, Danish Pruthi — Mon, 02 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 97%, Hallucination 52%

Given that AI agents can search the web broadly and deeply in real time, it is tempting to conclude that their outputs account for recent world events. We show that this conclusion does not hold in practice. While agents can search and easily find recent changes, the correctness of their responses depends on whether they search and what they search for. Popular agentic search benchmarks like BrowseComp cannot expose failures wherein agents do not recognize their own assumptions that should be verified, since these benchmarks explicitly state what needs to be found. In contrast, many user requests state only the final information need (e.g., an itinerary for a trip), leaving the agent to discover relevant changes (e.g., the temporary closure of a museum) and account for their consequences. To highlight such gaps, we introduce BlindToChange, a benchmark that measures whether agents proactively discover recent events that affect their responses. Our benchmark comprises requests for recommendations where an obvious candidate is no longer applicable due to a recent real-world event. These events are drawn automatically from Wikidata, allowing BlindToChange to be periodically refreshed. Strikingly, we find a big gap -- as high as 27.5 points, even at high reasoning effort -- in systems' performance, when the recent event is not explicitly mentioned compared to when we ask about it directly. Analyzing search trajectories, we learn that more than half of the failures can be attributed to the fact that agents do not directly search to verify whether a relevant fact has changed. Our benchmark highlights, and tracks, a new axis of progress, wherein systems account for recent information that is not explicitly mentioned in the user request.

★ Multi-Test Nation Bias Benchmarking of LLMs with Explicit Unbiased Baselines

Jonghyeon Choi, Yeonjun Choi, Beakcheol Jang — Mon, 02 Nov 2026 00:00:00 +0000

98% · Evaluation 98%

While prior work has highlighted that systematic nation-level bias in large language models (LLMs) can pose operational risks for international relations (IR) applications, many existing evaluations still lack a clearly specified unbiased reference (ground truth), limiting fully quantitative and cross-setting measurement. Using an expanded corpus of 722 official UN Security Council (UNSC) records, we measure nation-level bias with five complementary tests, thereby extending prior approaches to cover a broader set of bias expression channels: opinion ranking (WOW), association sorting (AK+-), vote simulation (PVS), opposition attribution (F-Oppose), and framing (Headliner). To enable grounded bias evaluation, we explicitly define an unbiased status for each test; in particular, for WOW and AK+- where ground truth is difficult to specify, we propose an action-based pseudo unbiased status derived from observable voting behavior to construct a quantifiable baseline. Benchmarking 4 LLMs (GPT-4o-mini, Llama 3.3-70B, Mistral-22B-Small, and Qwen3-30B) across the five permanent members of the UNSC, we find a test-agnostic macro-pattern: relative to the reference baselines defined for each test, model outputs lean toward the opposition (red) side for China and Russia and toward the supportive (blue) side for the U.K., the U.S., and France. These results underscore the need to audit IR-facing LLM deployments with grounded, multi-test benchmarks tied to explicit reference points.

★ TradeWar: A Benchmark for Execution-Grounded Trade Negotiation with Adaptive LLM Agents

Xiangyu Kong, Ruirui Yang, Xia Hou — Mon, 02 Nov 2026 00:00:00 +0000

98% · Evaluation 98%, New model 32%

Multi-turn negotiation remains a challenging problem for large language model (LLM) agents. While LLMs can conduct natural-language dialogue, existing benchmarks predominantly assess only agreement rates, overlooking whether negotiated commitments actually create value once executed. To address this gap, we introduce TradeWar, a multi-country simulation benchmark that translates negotiation dialogues into concrete tariff and trade-share policies, enforces capacity and network constraints during settlement, and evaluates outcomes via an Economic Health Index (EHI). To solve this execution-grounded task, we propose TACTIC (Trade-aware Adaptive Consolidation and Theory-of-mind for International Commerce), an agent that iteratively refines its negotiation strategy across episodes-without updating base LLM weights-by leveraging prior experience, opponent behavior, and past settlement outcomes. Despite using zero-shot chain-of-thought (CoT) prompting, PromptLLM attains a 75% agreement rate but only 57.1 final EHI (*100), whereas TACTIC reaches 68.8. Across 20 sequential episodes and three seeds, TACTIC's final EHI improves from 58.2 to 73.2, compared with 65.2 for an ablation without cross-episode updates. Under technology-controls and energy-crisis shocks, TACTIC also exhibits smaller performance degradation than PromptLLM and MemoryLLM. These results indicate that CoT deliberation and interaction memory alone do not fully address delayed execution feedback, whereas settlement-grounded adaptation improves long-horizon negotiation in TradeWar.

★ Mage-VL-Cowork: A Full Training Recipe to Learn Hybrid Computer Use From Evolved Harnesses

Xiaoyi Zhang, Haoqing Wang, Xinjie Zhang, Xiang Ming, Xing Wang, Senqiao Yang — Mon, 02 Nov 2026 00:00:00 +0000

97% · Agents 97%, New model 94%, Efficiency 79%

Effective computer-use agents (CUAs) should seamlessly combine visual GUI interaction with code and terminal execution within a single policy. We present Mage-VL-Cowork, a 4B hybrid CUA trained from a general multimodally aligned checkpoint through an end-to-end, progressive five-stage curriculum across midtraining, SFT, and RL. To power this curriculum at scale, Mage-CUA-Forge replaces per-task agent construction with reusable, skill-conditioned generation workflows and application-specific preparers, synthesizing 293,806 desktop and web tasks across 6,915 skills while reducing recurring per-task expenditure by 82.6%. Rather than directly distilling a default teacher, we evolve the teacher harness and instantiate specialized CUA harnesses to yield high-quality trajectories alongside stage-specific supervision. Mage-VL-Cowork-4B finally climbs to 66.0% on OSWorld-Verified through different training stages with more efficient reasoning cost, achieving state-of-the-art performance at its parameter size and pushing forward the frontier of on-device intelligence.

★ Does Recursive Knowledge Construction Help Agents Verify Consensus Claims in Science?

Mon, 02 Nov 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 81%, Hallucination 35%

Recent scientific agents build knowledge recursively by retrieving evidence, constructing an intermediate state, and judging whether that state contains sufficient information. These agents are evaluated end-to-end, usually without ground truth, so the effect of each design choice is unknown. Here, we examined two design axes: (1) what the recursion builds (the state of evidence) and (2) what controls it (the judge that decides when to stop). To measure them, we defined consensus scientific claim verification and measured these effects on this task, where a system must establish a claim's consensus stance from the open literature. We built a new consensus claim dataset, ProClaim-eval, including 718 claims with SUPPORT and REFUTE gold labels from five expert-curated databases spanning molecular biology, nuclear physics, and astronomy. Our instrument, ProClaim, persists an evidence state outside the model's context, so each axis can be studied with the other held fixed. Against verification systems and general-purpose agent harnesses, ProClaim ranks among the top three, with a significant advantage over the harnesses. We found that the evidence state improves performance over in-context accumulation, and it also matters what is stored in it. In addition, a trained judge that reads the evidence state performs better than fixed-round stopping. A judge trained separately and never exposed to ProClaim-eval achieves the highest agreement with the gold labels on three of five subsets, where it matches the off-the-shelf Claude Sonnet~4.6 judge and outperforms Claude Haiku~4.5. Our results support recursively maintaining a structured evidence state outside the model's context as an effective design for verifying claims in scientific agents.

★ Capability-Grounded Agentic Planning

Shikha Shrestha, William Yeoh, Ning Zhang, Yevgeniy Vorobeychik — Mon, 02 Nov 2026 00:00:00 +0000

97% · Agents 97%, Hallucination 56%, Evaluation 42%

LLM agents serve natural-language requests through tools, data sources, and other agents, and even simple requests often need several interdependent calls. Typically, the LLM picks these calls by trial and error, with no guarantee of an executable plan. We present Capability-Grounded Agentic Planning (CGAP), which instead reasons about what a request requires and what tools provide. CGAP compiles the tool catalog deterministically into a PDDL model, uses an LLM to turn the request into goals, and solves the model with a classical planner. Since tool outputs are unknown in advance, the model is abstract, tracking which inputs are available, not their values. During execution, an LLM adds entity-specific calls that abstract plans omit, and deterministic checks reject any that break the plan. Every argument is drawn from a cited source, such as a conversation span or an earlier tool response, and each requested result counts as delivered only after an LLM verifier checks it against the request. With three LLMs on ToolSandbox, ComplexFuncBench, BFCL multi-turn, and $\tau^2$-bench retail, CGAP outperforms a standard tool-calling agent by up to 28.9 points.

★ From Transcription to Reformulation: Defining and Benchmarking Grounded Generative Speech Recognition

Qing Wang, Qihan Hu, Hang Lv, Yan Yang, Boqing Liu, Xuechao Zhu — Mon, 02 Nov 2026 00:00:00 +0000

97% · Evaluation 97%, Hallucination 81%

Speech language models have expanded automatic speech recognition (ASR) from transcription to general speech-to-text generation. Yet existing benchmarks still assume a single ground-truth transcript per utterance, an assumption that struggles when models produce readable, formatted, or structured outputs. In this work, we introduce Grounded Generative Speech Recognition (GSR), reformulating speech recognition as generating semantically grounded text instead of a single transcript. We define a grounded reformulation space that characterizes all valid textual realizations supported by the input speech, and establish a four-level hierarchy of reformulation capabilities, including literal transcription, canonical reformulation, semantic reformulation and speech-grounded structuring. Based on this formulation, we present GSR-Bench, the first comprehensive benchmark designed to evaluate generative speech recognition systems beyond exact transcript matching. GSR-Bench covers diverse scenarios including refined meeting transcription, cross-dialect translation, context-aware speech understanding, and robustness under silent or noisy acoustic conditions. To address the limitations of word-level evaluation metrics, we develop a multi-dimensional evaluation framework combining speech grounding assessment and LLM-based semantic judgment, whose reliability is validated through human preference studies. Extensive experiments across representative architectures, ranging from conventional ASR systems to end-to-end Speech-LLMs, reveal fundamental trade-offs between transcription fidelity and generative reformulation ability. Our analysis further identifies critical failure modes, including acoustic hallucination, unsupported generation, and instruction drift, providing insights into the design and evaluation of future grounded generative speech systems.

★ StrictPRMBench: A Replayable Evaluation Benchmark for Auditing Fixed-Score PRM Aggregation

Pengfei Zheng — Mon, 02 Nov 2026 00:00:00 +0000

97% · Evaluation 97%

Process reward models (PRMs) score reasoning steps, but best-of-N selection also requires a rule that aggregates those scores. We study that rule as a separate evaluation object. StrictPRMBench fixes mathematical reasoning pools, scorer evidence, and the source trace of a deletion intervention, so that different reducers face the same recorded selection problem. Across seven scoring checkpoints and two benchmarks, the stored N=32 outcomes give higher deletion-induced selection instability for product than geometric mean (GM) in all 14 cells. The equal-cell difference is +0.0965; a shared-problem bootstrap, conditional on these checkpoints and frozen pools, gives a 95% interval of [0.0846,0.1086]. A pre-injection-defined sensitivity excludes replacement slots that are either reducer's clean winner and retains 4,459 of 4,622 cell-problem records. The retained difference is +0.0986 [0.0875,0.1098]. Effects are heterogeneous: the two Qwen cells have positive percentile-bootstrap intervals on retained support but two-sided exact paired p-values of 0.0625 and 0.0703. Clean-pool Qwen accuracy does not detect a product advantage, but this is not an equivalence result. We provide versioned outcome records, CPU reanalysis, and an explicit evidence-availability contract. The contribution is a diagnostic evaluation protocol and a reproducible audit of recorded selector behavior, not a new PRM, a universally optimal reducer, or evidence of deployment-wide harm. Raw-score replay and outcome-record reanalysis are distinguished: missing primary edited-score vectors limit independent evaluation of new reducers under the same stress.

★ Do Transaction Prices Reflect Visual Preference? A Diagnostic Study with Multimodal Language Models

Fengxian Ji, Jingpu Yang, Fan Zhang, Junhong Liang, Qian Jiang, Yuke Li — Mon, 02 Nov 2026 00:00:00 +0000

97% · Evaluation 97%

Existing visual preference evaluation relies on costly human annotations or preference reward models, whereas online marketplace transaction prices are readily available and reflect real purchasing behavior, offering potential low-cost supervision. However, transaction prices conflate visual preferences with non-visual factors such as provenance, demand, and competition, making aggregate accuracy insufficient to distinguish model errors, unreliable price signals, and protocol confounds. To address this identifiability challenge, we introduce MarketImage, a diagnostic framework that identifies when transaction prices reflect visual preferences, where they fail, and whether multimodal language models can recover human-supported visual preference signals. MarketImage comprises three components: (i) MarketImageArt, 266 six-image comparison sets with transaction prices and price-blind human rankings; (ii) MarketImageLens, a tie-preserving protocol that partitions 3,990 within-set image pairs into price-human agreement, conflict, and price-tie regions while separately auditing presentation-order effects; and (iii) MarketValuator, six elicitation methods spanning direct comparison, independent scoring, budget constraints, and multi-agent aggregation to evaluate signal recovery by static visual scorers and three multimodal language models. Experiments show that only 24.2% of image pairs have a non-tied price direction, of which 34.7% agree with the human majority. Within this region, HPSv2 leads at 66.4% Recoverable Accuracy; Qwen's budgeted and aggregated variants trail baselines, while GPT interface gains remain statistically unresolved. Thus, transaction prices encode recoverable visual preferences only under limited conditions and should be treated as uncertain proxies rather than ground truth.

★ RTL-Interact: Training EDA-Interactive Models for Multi-Task RTL Design

Wenxuan Shi, Yaoyu Zhu, Di Huang, Chongxiao Li, Xiaoyun Zhang, Muxin Song — Mon, 02 Nov 2026 00:00:00 +0000

96% · New model 96%, Agents 76%, Efficiency 71%

Large language models offer a promising route to register-transfer-level (RTL) design, but reliable implementations require checking and revision based on electronic design automation (EDA) tools' feedback. Existing RTL-specialized models often lack learned interaction with such EDA tools or rely on separate generation and repair components. We introduce RTL-Interact, a multi-task post-training framework that trains a single RTL-specialized model to generate RTL, construct task-appropriate EDA checks, and revise implementations using tool feedback. The framework combines automatically constructed multi-task demonstrations with staged supervision from chain-of-thought to tool-integrated reasoning. We further introduce Pattern-Aware Policy Optimization (PAPO) for multi-task reinforcement learning. PAPO augments Group Relative Policy Optimization with correctness-gated rewards that favor concise successful checking while rewarding effective revision, and uses task grouping to preserve coverage across objectives. On CVDP, RTL-Interact-32B achieves 48.7% weighted Pass@1 and 68.9% Pass@5, surpassing its teacher model, DeepSeek-V3.2, while RTL-Interact-8B achieves 44.9% weighted Pass@1. Additionally, we show that PAPO can reduce average tool calls by 63.8% while improving Pass@1. Both models also outperform all RTL-specialized baselines on VerilogEval v2 and RTLLM v2, demonstrating the effectiveness of our method.

★ LEAP: Efficient LLM Inference Scheduling with Generalizable Length Prediction

Kwan Yan, Jialun Zhong, Yanzeng Li — Mon, 02 Nov 2026 00:00:00 +0000

96% · Efficiency 96%

Length-aware scheduling reduces request latency in LLM serving by prioritizing shorter requests. However, despite their large model sizes, length predictors can suffer degraded ranking performance under domain and task shifts, and adapting them can require costly retraining. Their prediction overhead can further offset the benefits of scheduling. To improve the generalization and efficiency of length-aware scheduling, we propose \textbf{LEAP}, a length prediction system with fast online adaptation and asynchronous CPU execution. LEAP reuses the serving model's prefill hidden states, which contain information about the output length, and extracts this signal through a linear readout. To adapt to substantial domain and task shifts, it updates the readout at test time through recursive least squares, using completed requests as supervision without gradient computation or historical-data replay. Prediction and updates run asynchronously on CPU, while prediction and initial batch ordering overlap the remaining GPU prefill. Across the evaluated cross-domain transfers, LEAP retains strong ranking performance, and online adaptation improves Spearman correlation by 64.7\% under a larger domain shift. Serving experiments show that LEAP reduces average per-token latency by 15.9--40.8\% compared with the evaluated length predictors, while achieving 1.9\% higher throughput with $30{,}569\times$ fewer predictor parameters than the strongest evaluated baseline.

★ UpLoop: Training Looped Transformers via Upcycling at No Extra Pretraining Compute

Cheng Liang, Tengfei Zhang, Ya Zhang, Yanfeng Wang, Zhouhan Lin — Mon, 02 Nov 2026 00:00:00 +0000

96% · New model 96%

Looped language models can strengthen reasoning through repeated computation with shared weights, increasing execution depth without adding independent parameters. However, each additional pass increases training floating-point operations (FLOPs), making these gains more expensive to obtain. To address this cost, we propose UpLoop, a pretraining schedule that uses upcycling to train looped Transformers while exactly matching a standard dense baseline in final parameter count, total training tokens, and training FLOPs. Specifically, UpLoop begins with a shallow Transformer, progressively grows its depth by copying learned layers, and uses the computation saved early to support repeated execution of a shared middle block later in training. Experiments across model scales show improvements in language modeling and downstream performance, with the average score across 22 downstream tasks increasing from 22.99 to 27.52 at 1.3B parameters. These gains also extend to mathematical reasoning after supervised fine-tuning.

★ Latent Sparse Attention: Towards Subquadratic Ultra-Long Context Attention

Mon, 02 Nov 2026 00:00:00 +0000

96% · Efficiency 96%, New model 78%

Agentic workflows are increasingly shifting towards long-horizon tasks, such as open-ended scientific work, where models benefit from long context windows. However, at ultra-long context (above 1M tokens), full attention is prohibitively expensive. A promising approach is sparse attention, in which a dedicated indexer selects a subset of keys for the attention. Yet current methods of indexing either still have quadratic cost in context size or lose accuracy at ultra-long context scale. We address this with Latent Sparse Attention (LSA), an efficient attention method that maintains accuracy at ultra-long context. In LSA, keys are hierarchically clustered in a learned latent space, and queries score the hierarchy of clusters instead of scoring every key, making prefill, decode and training faster at long context. We propose two variants, Distilled-LSA (D-LSA), which can be applied to pre-trained models via distillation, and Native-LSA (N-LSA), which is trained natively on the next-token loss alone. On GLM-5.3-FLASH (QWEN-3.8-27B) at 10M tokens, D-LSA speeds up one attention layer by 22.7x (72.4x) in prefill and 4.6x (42.5x) in decode over its original sparse attention (full attention), while keeping similar accuracy across a variety of tasks. N-LSA matches the accuracy of full attention with a training speedup of 1.6x per layer at 128k tokens. LSA offers practical and efficient attention for models to benefit from ultra-long context.

★ WeVisCut: The Whole Editing Room in One Agent From Narration to Long-Form Video

Chenyu Mu, Wenming Weng, Qinrui Wang, Xingsong Ye, Hao Yu, Jun He — Mon, 02 Nov 2026 00:00:00 +0000

96% · Evaluation 96%, Agents 92%, Efficiency 85%

Producing long-form videos from narration requires interdependent decisions that remain coherent across shots. Existing agentic editors often use multi-stage workflows driven by general models, leaving the quality-cost trade-off of compact, task-adapted policies unclear. We address this gap with WeVisCut, a long-form script-to-video system that connects tool execution with policy learning. Its agent harness maintains shared production state so that tool feedback informs subsequent decisions and repairs. Because finished videos do not expose these decisions, our four-stage data pipeline combines editing inversion with verified replay to turn recovered editing intent into executable supervision. Applied to 3,750 finished videos, this pipeline yields over 130K training examples. We use this supervision to train a single 9B policy that coordinates production and recovers from failures within the same harness. Moreover, we introduce a 100-task benchmark and five metrics assessing editing quality and delivery reliability. Under a shared tool environment, WeVisCut substantially outperforms its base model and achieves comparable performance to Claude Opus 5 while using nearly 33x fewer generated tokens per successful output minute. It also outperforms task-matched native editors on our proposed metrics.

★ FARO: Function-Space Adaptive Return-to-Risk Optimization via Cross-Block Equal-Risk-Contribution Budget Allocation

Xiaojun Wu, Cehao Yang, Xuhui Jiang, Chengjin Xu, Jia Li, Jian Guo — Mon, 02 Nov 2026 00:00:00 +0000

95% · New model 95%, Evaluation 37%

Training a large language model is compute-bound, and current recipes divide the work between optimizers by hand: orthogonalized momentum for attention and MLP matrices, Adam-style updates for the rest. This division rests on human priors about which layer type deserves which geometry, and it fixes a single global update budget that every parameter block shares. We introduce FARO, a function-space adaptive optimizer that makes the partition of parameters into blocks the unit of decision. FARO gives each block a geometry by its realized counterfactual return, then freezes the routes and reallocates the frozen per-block budget by cross-block equal risk contribution (ERC), the risk-parity principle from portfolio theory. On a GPT-style model trained on FineWeb, the framework gives two results. First, its FARO-Adaptive arm reaches 3.2780 nats at the full schedule, matching NorMuonAndAdam (3.2794), the best hand-designed recipe: it recovers the human prior without hand-setting per-layer geometry. Second, FARO-ERC improves on that arm: it lowers the clean-data loss at every one of 6 seeds, by 0.0136 nats on average, a modest but robust gain. Its sign test, paired test, and confidence interval all agree, and it exceeds the measured run-to-run noise at no added time or memory. It works by trimming the riskiest blocks rather than expanding the safest, and it holds up under a corrupted stream, still lowering the mean loss and winning the seed majority. The source code is available at https://github.com/wxj630/FARO.

★ Counterfactual Pairwise Augmentation Improves Small Pointwise LLM Judges

Mohammad Zangooei, Peyvand Teymoori, Mohammad Robati Shirzad, Soroosh Shahtalebi — Mon, 02 Nov 2026 00:00:00 +0000

95% · New model 95%, Evaluation 45%

Agentic systems increasingly gate individual steps with a pointwise judge that scores a response against a criterion. Small language models are attractive for this role because they can be served and specialized efficiently. Such a judge must locate the criterion's decision boundary, but pointwise labels under-determine that boundary in fine-tuning: labeled responses to different queries differ along many dimensions at once, so they do not reveal which features decide the label. We introduce CPA (Counterfactual Pairwise Augmentation), which makes the deciding difference explicit in training. For each pointwise item, CPA synthesizes and audits a counterfactual response to the same query intended to carry the opposite label. It frames the accepted pair as a pairwise task asking which of the two responses satisfies the criterion. The judge is fine-tuned jointly on both pointwise and pairwise datasets and deployed in pointwise mode only. A second-order analysis shows when this should outperform the prevailing strategy of merging separately trained pointwise and pairwise experts: merging moves every parameter direction toward the pairwise expert by the same fraction, whereas joint training can hold the pointwise optimum wherever the pairwise loss is flat and absorb its curvature where it is sharp. This is the regime that CPA's same-source, counterfactually minimal pairwise view is designed to produce. On separate Qwen3-8B LoRA adapters for six binary criteria, CPA improves criterion-averaged pointwise macro-F1 by 2.5 points over pointwise-only training and outperforms model merging on all six criteria at every tested merge weight. It also surpasses the strongest zero-shot frontier judge we evaluate, Gemini 3.6 Flash, by 11.8 points.

★ TraceSpec: Understanding and Improving Speculative Decoding for LLM Agents

Jihao Zhao, Shuaishuai Zu, Pengfei Li, Biao Qin, Guanghuan Fang, Ben Feng — Mon, 02 Nov 2026 00:00:00 +0000

95% · Agents 95%, Evaluation 92%, Efficiency 89%

LLM agents execute tasks through repeated model calls over evolving histories of reasoning, actions, and tool feedback. Aggregate speculative-decoding metrics merge these distinct decisions, providing limited guidance on where draft supervision is needed. We present TraceSpec, a framework connecting trajectory evaluation with adaptive draft training. It combines dynamic execution with fixed-prefix replay and introduces TraceSpec-Traj, a corpus of 10,000 training states and 2,500 test states spanning browsing and software engineering. Analysis of five speculative baselines reveals that aggregate acceptance can rise with context length while declining within matched late browsing actions. Workload and phase further distinguish these trends, motivating supervision guided by trajectory conditions. TraceSpec trains on intermediate assistant turns and allocates a fixed prediction budget using state-conditioned estimates of prefix-aware acceptance gaps. Across three target sizes and two temperatures, it improves replay throughput by 12.5% to 22.6% over its DSpark base. Qwen3-4B ablations show that adaptive allocation increases accepted length by 3.3% to 10.0% over fixed full-turn supervision, with additional comparisons substantiating the roles of state conditioning and consecutive-token agreement. The resulting evaluation resources, trajectory findings, and training strategy connect the analysis of agent execution to practical draft improvement.

★ In-Model Context Compression with Agentic Soft Tokens for Long-Horizon LLM Agents

Xuning He, Kai Fan, Runyu Zhang, Binghuan Wang, Yuxiang Zhang, Xue Jiang — Mon, 02 Nov 2026 00:00:00 +0000

95% · Efficiency 95%, Agents 32%, Evaluation 30%

Long-horizon LLM agents play an important role in solving complex and sustained tasks, with ever-growing interaction contexts. However, these accumulated contexts will raise substantial computational cost and obscure key decision-relevant evidence. Existing methods typically train external compressors or context-management agents to rewrite context as shorter hard-token sequences, but such external selection may not align with the LLM's evolving representation and is necessarily lossy. In this paper, we introduce Agentic Soft-Token Compression (ASTC), a training-free compression method using soft tokens within LLM agent for long-horizon tasks. Specifically, ASTC uses the model's native recurrent dynamics to determine the already absorbed information, leveraging the commonly-used hybrid Gated Delta Network (GDN)-attention. It then carries this information forward as agentic soft tokens while retaining exact values as hard tokens. The resulting compact sequence reduces later-layer computation and yields reusable KV caches, extending the savings across the agent's trajectory. Extensive experiments on long-horizon tasks show that ASTC matches or exceeds the uncompressed agent's accuracy with up to 50.68% fewer end-to-end Attention Pairs, while trained compressors require 15.4 to 103.7 times more Attention Pairs per successful task. These findings highlight agentic soft tokens as a model-native way to carry absorbed context forward as the task representation evolves.

★ RECITE: Revising Eligibility Criteria Incorporating Textual Evidence

Russell Ro, Reza Abbasi-Asl — Mon, 02 Nov 2026 00:00:00 +0000

95% · Evaluation 95%, Agents 62%, Hallucination 42%

Overly restrictive eligibility criteria contribute to widespread enrollment failure in clinical trials, yet published recommendations for how specific criteria should be broadened remain disconnected from trial design at scale. We introduce RECITE (Revising Eligibility Criteria Incorporating Textual Evidence), a framework for automated discovery, extraction, and implementation of evidence-based eligibility criteria revisions. RECITE uses protocol amendment documents from ClinicalTrials.gov to construct a dataset of 11,913 curated instances across 5,735 trials, each pairing original criteria with documentary evidence and target revisions. We benchmark fifteen LLM configurations under truncated-to-context-window and retrieval-augmented generation conditions, on a 3,116-instance sample under an LLM-as-judge protocol validated against ten domain experts (judge vs. human quadratic-weighted Cohen's $\kappa_{qw} = 0.694 \pm 0.096$). The best open-weight model, Gemma 2 9B, achieves $\kappa_{qw}$-aligned ordinal score $3.0 \pm 0.9$ on a 0--4 scale (normalized 0.742) and the best closed-weight model, GPT-4o, reaches $3.0 \pm 0.8$ (normalized 0.745). Using an agentic literature discovery system, we then identified 44 paper--trial matches across five publications providing trial-design and EC-modification evidence. Our findings suggest that existing clinical knowledge can be operationalized into actionable trial design recommendations.

★ BYOD: Build Your Own DLM - Efficient Autoregressive-to-Diffusion Language Model Conversion

Ruurd J. A. Kuiper, Robert A. Bagheri — Mon, 02 Nov 2026 00:00:00 +0000

95% · New model 95%, Evaluation 63%, Efficiency 47%

Unlike autoregressive (AR) models, diffusion language models (DLMs) permit parallel token updates, flexible generation, and bidirectional attention, but leading systems rely on costly pretraining or full-model fine-tuning. We present Build Your Own DLM (BYOD), a generalizable LoRA-only recipe for parameter-efficient AR-to-DLM conversion. Without architecture-specific tuning, BYOD induces diffusion generation in Gemma 2 9B, Llama 3.1 8B, Qwen2.5 7B, and Ministral 8B; to our knowledge, this is the first reported AR-to-DLM conversion of a Mistral model. The adapters comprise 4.1-5.8% of final parameters while every base-model parameter remains frozen; each 25,000-iteration conversion needs at most 102.4M tokens, processed in 3.1-4.0 hours on one Google Colab GPU. All four models generate from fully masked sequences, use bidirectional context in a controlled attention test, and retain substantial parent-model knowledge. BYOD-Gemma is the strongest BYOD conversion on our nine-task evaluation, reaching 50.6% average accuracy compared with 51.9% for LLaDA 8B Instruct in the same harness. Open-generation quality at low network function evaluation (NFE) budgets remains the principal limitation: BYOD perplexity is 8.3-11.1 versus 4.1 for LLaDA and 2.2-3.0 for the AR parents when generating 128 tokens in 64 steps. To enable further experimentation and inspection, we provide a standalone training and evaluation implementation and interactive demos for all four conversions. BYOD thus provides a practical route for extending diffusion generation to other existing model families while identifying the gap that efficient conversion must next close.

★ ACCT: Asymmetric Cross-LLM Context Transfer

Yuzhe Liu, Haolin Tian, Tonghan Wang — Mon, 02 Nov 2026 00:00:00 +0000

95% · Efficiency 95%, Evaluation 49%

In real-world deployments, large language model (LLM) systems are increasingly evolving from single-model invocation toward heterogeneous deployments in which different models are dynamically invoked for inference. When the same context is reused across models, however, each newly selected target model typically has to process it again, repeatedly incurring substantial computational costs. Therefore, we propose ACCT (Asymmetric Cross-LLM Context Transfer), which targets KV cache transfer between models of different scales within the same model family. ACCT transforms the KV cache produced by a source model into sparse KV states that can be directly consumed by a target model, thereby enabling cross-model reuse of context computation and decoupling context computation from downstream inference. ACCT consists of three lightweight components: the Cartographer identifies query-relevant context tokens for each target layer and KV head; the Allocator determines how many context tokens each target layer and KV head retains under a shared total KV budget; and the Compiler transforms selected source KV into sparse target KV. The target model can therefore perform inference using the transferred KV without reprocessing the original context. We evaluate ACCT in two complementary settings comprising five general-capability benchmarks and six highly context-dependent tasks. Experiments show that ACCT matches or exceeds the target model's original inference performance even when reducing the KV cache size by 75%. At a context length of approximately 19K tokens, ACCT's KV conversion achieves a 6.5-21.2 times speedup over target context prefill. These results demonstrate the feasibility of transferring context computation across models under a limited KV budget.

Evidence-Guided Recovery for AI Agents with Ambiguous Tool Outcomes and Partial Execution

Genliang Zhu, Xiaobo Chen, Ziyuan Wang, Chu Wang — Mon, 02 Nov 2026 00:00:00 +0000

94% · Agents 94%, Evaluation 81%, Efficiency 30%

Tool timeouts can leave AI agents facing ambiguous tool outcomes and partial execution: an action may have failed, committed or remained in flight while other workflow steps have already completed. We formulate evidence-guided task recovery over approved obligations, provider capabilities and states compatible with public evidence. A shared admissibility check handles key expiry, authorization timing, dependencies and residual effects; a terminal certificate separates actual goal satisfaction from publicly supported safe completion. The known-model implementation combines standard observation-contingent planning with a separately implemented effect-history adjudicator. In a finite study of 37 assignments, planning and three active controls each certify 28 completions; four saved synthetic fee units incur additional computation. A preregistered assignment-held-out comparison then evaluates planning advice against evidence prompting with two local model bundles. Across 58 paired units in 19 previously developed task families, advice certifies 30 completions versus 9, a 36.2-percentage-point difference (exact family sign-flip p=0.0078), using 272 proposals versus 399. All 116 trajectories pass independent replay. Both models exhibit a positive completion difference. The comparison measures the evaluated advice interface under a common shield; search-versus-rule advice requires a separate matched comparison. Unknown and handoff outcomes remain explicit, with guarantees tied to the declared provider capabilities rather than arbitrary third-party exactly-once execution.

EvoMCP: Co-evolving MCP agents and training data

Xuanang Gao, Kepeng Lin, Xinyan Shi, Yong Deng, Shuheng Shen, Changhua Meng — Mon, 02 Nov 2026 00:00:00 +0000

94% · Agents 94%, New model 92%, Evaluation 38%

Supervised fine-tuning of Model Context Protocol (MCP) agents is typically conducted on a static data distribution, which fails to adapt to the student's evolving capabilities and leaves newly exposed weaknesses insufficiently addressed. To address these challenges, we propose \method, an agent--data mutual evolution framework that enables MCP agents to improve through a closed loop of task evolution, trajectory evolution, and supervised fine-tuning. Specifically, \method\ assesses the student's established capabilities and capability gaps from rollout feedback to identify learning targets. Task evolution then translates these targets into executable tasks by adapting the instructions and observable environment states of compatible verified seed tasks while keeping their verifiers aligned. To provide reliable and efficient demonstrations for these tasks, trajectory evolution uses reusable skills to recover failed teacher rollouts and refine successful ones. The resulting task--trajectory pairs are used to fine-tune the student, while accepted tasks seed subsequent evolution. To close the loop, we re-evaluate the updated student and use its feedback to guide the next round of data evolution. Experiments on Toolathlon, MCPMark, and MCP-Atlas demonstrate consistent improvements over the initial student, yielding absolute gains of 16.5\%, 10.2\%, and 12.6\%, respectively. Project page: https://github.com/xagao/EvoMCP.

ReX-GG: a LLM Ensemble Pipeline for Relation-extraction and Graph Generation

Giacomo Magnifico, Eduard Barbu — Mon, 02 Nov 2026 00:00:00 +0000

94% · Hallucination 94%, Evaluation 78%, Efficiency 52%

Current LLM ensemble frameworks focus on multi-step setups with additional modules for answer ranking, often opting for token and span analysis rather than structured outputs, leading to heavyweight architectures with potential fail states along the pipeline. Faster, lighter solutions are more vulnerable to hallucination propagation and can lack output control in more complex pipelines. This paper proposes a customisable, lightweight ensemble workflow of coordinated Large Language Models that leverages JSON-structured outputs and anonymous peer-review ranking to mitigate hallucinatory outputs and single-model failure points. The pipeline is demonstrated on a relation extraction task applied to popular science articles in English, targeting four ontologically-grounded relation types (strong causation, weak causation, contrastive, and compositional), with semantic node canonicalisation and interactive, colour-coded HTML causal graphs as the final output. Performance is evaluated through an anonymous user study, achieving an average perceived accuracy of 0.778 against a human-annotated gold standard. The modular architecture supports flexible deployment across both API-based and in-house LLM setups, and the full framework is released under an open license to foster reproducibility and collaborative research.

Asynchronous Flow Map Language Models

Mon, 02 Nov 2026 00:00:00 +0000

94% · New model 94%, Efficiency 83%, Evaluation 31%

The order in which tokens are predicted in language modeling has a profound impact on the quality of generated sequences. Autoregressive and masked diffusion models iteratively fill in tokens and can exploit such orderings across generation steps, but how to leverage an ordering for one-step parallel language modeling is unclear. We introduce asynchronous flow map language models (aFMLMs), which incorporate prediction order into one-step modeling by learning the flow maps of asynchronous flows. These flows specify an ordering through a tokenwise time schedule that controls the relative denoising progress of tokens. We formulate a general class of asynchronous flows that admit both prescribed and learned, state-adaptive schedules, show that they define valid flow maps, and develop new methods for their flow map distillation. The resulting models can generate all tokens in a single forward pass by internalizing the ordering encoded in the flow into their hidden computations. On Sudoku solving and code generation for GSM8K math word problems, aFMLM establishes new state-of-the-art one-step performance.

Subquadratic Attention with a Hierarchical Tree-Structured Indexer

Mon, 02 Nov 2026 00:00:00 +0000

94% · Efficiency 94%, New model 52%

The quadratic scaling of attention is a fundamental bottleneck limiting the long-context capabilities of large language models (LLMs). A variety of approaches have attempted to circumvent this, but they either fail to match attention's expressiveness (e.g., sliding-window or linear attention variants with fixed-size state) or only achieve constant-factor speedups (e.g., sparse attention methods whose indexers themselves scale quadratically). In this paper, we derive an asymptotically subquadratic sparse attention method based on a hierarchical, tree-structured indexer, which we call Zyphra Hierarchical Attention (ZHA). ZHA constructs a tree of recursively compressed key summaries over the KV cache, which future queries traverse from the top down to retrieve a sparse subset of keys. Crucially, by iteratively growing the tree's depth, the total cost including routing and final attention grows sublinearly in sequence length, resulting in overall subquadratic (but superlinear) scaling for attention. We demonstrate that ZHA can be cheaply applied to existing pretrained models with only a short KL distillation training stage over a few billion tokens, and maintains performance on long-context retrieval and challenging reasoning, mathematical, and agentic coding evaluation tasks. Moreover, we show that our method is highly robust to tree geometry including very deep trees of up to 6 levels. Finally, we develop a detailed performance model and show that our approach scales subquadratically, dramatically outperforms competing methods beyond 1 million context, and ultimately has the potential to extend effective context windows to the hundreds of millions or billions of tokens.

DecisionRank: On-Policy Distillation for Decision-Model Rerankers

Yanran Wu, Dian Yu, Chongyang Bai, Jieyi Zhang, Rui Liu, Tianyi Chen — Mon, 02 Nov 2026 00:00:00 +0000

94% · New model 94%, Efficiency 93%, Evaluation 36%

Decision models target structured tasks by directly predicting choices or scores, avoiding autoregressive decoding latency. However, transferring a generative LLM's ranking knowledge to a decision model remains challenging: the teacher expresses preferences through a ranked list of document ID sequences, while the student determines the ranking using scores from the first-token position. To bridge this gap, we introduce DecisionRank, a decision model trained through item-space on-policy distillation (OPD). Item-space OPD constructs the student's rollout by sorting its document scores into a ranked sequence. A generative teacher takes the resulting ranking list as inputs and provides conditional feedback to directly supervise those scores. Combined with offline ranking supervision, DecisionRank achieves state-of-the-art performance on both general and reasoning-intensive retrieval benchmarks such as TREC DL, BEIR, and BRIGHT. Across four benchmarks, item-space OPD consistently outperforms alternative training objectives. Even at 1.7B parameters, DecisionRank surpasses its 27B generative teacher on TREC DL19--23. At 8B, it outperforms its generative listwise counterpart while running $24.5\times$ faster on TREC DL and $16.7\times$ faster on BRIGHT.

Flide: Accelerating Decision Inference with Exact State Forking

Bobo Li, Jiajia Song, Hao Fei, Wynne Hsu, Mong Li Lee — Mon, 02 Nov 2026 00:00:00 +0000

94% · Efficiency 94%, Evaluation 43%

Decision models such as Jev have emerged as a highly promising paradigm for LLM agent systems, providing lightweight and reliable categorical reasoning for rapid System One execution. However, existing generation-oriented inference engines fail to support these workloads effectively, suffering from severe prefix recomputation overheads or distorted output probabilities. To address this challenge, we propose Flide, an efficient and exact inference engine that decouples the logical tree structure of a decision request from its physical execution within a single flattened token stream. Through ForkAttention, Flide navigates request dependencies using GPU-resident indices across both standard attention and recurrent Gated DeltaNet layers, guaranteeing in exact arithmetic that every query branch evaluates identically to an isolated forward pass. By expressing branch topologies directly in GPU-resident metadata, Flide accommodates diverse request mixtures with a compact set of pre-recorded execution graphs while enabling expanding agent histories to resume from checkpoints saved in earlier passes. Extensive evaluations on public hybrid and decision models demonstrate that Flide answers queries up to 9.5 times faster than vLLM and SGLang while staying closest to a float32 reference in total variation distance. By unifying physical execution throughput with mathematical exactness, Flide establishes a principled foundation for scalable, latency-critical decision intelligence in agent systems.

Li$^{2}$SAO: Lifting Light Singl-Rollout Asynchronous RL Optimization via Group-Relative Actor-Critic Bootstrapping

Zichen Bai, Yanli Wang, Yang Liu, Junhao Shi, Xueqin Liu, Jinguo Sun — Mon, 02 Nov 2026 00:00:00 +0000

93% · New model 93%

Asynchronous reinforcement learning (RL) is becoming a mainstream paradigm for post-training large language models (LLMs), as it improves training efficiency by updating the model as rollouts arrive. Recent work argues that group-wise sampling in GRPO does not naturally fit asynchronous RL training because rollout lengths follow a long-tailed distribution, and proposes Single-rollout Asynchronous Optimization (SAO) to improve stability and off-policy behavior, especially on agentic tasks. However, SAO relies on a closed-source value model obtained from scaling value pretraining. In this paper, we analyze GRPO and LiSAO, our lightweight reproduction of SAO without the pretrained value model, on single-turn reasoning and multi-turn agentic tasks. We find that GRPO outperforms LiSAO on reasoning tasks, while the reverse holds on agentic tasks. Since the value model of LiSAO is not yet well trained , LiSAO lags behind GRPO substantially in early training steps. To address this issue, we propose Li$^{2}$SAO, which Lifts Light Single-rollout Asynchronous Optimization with group-relative actor-critic bootstrapping: while the critic is cold, a group-relative advantage trains the actor and a group-centered return trains the critic, after which training switches to single-rollout GAE. Experiments show that Li$^{2}$SAO catches up with or surpasses GRPO in early training and further raises the later performance ceiling. Code is available at https://github.com/jd-cc-coder/Li2SAO.

Rome Wasn't Built in One Turn: Replay-Grounded Reward Signals for Long-Horizon Freeciv Agents

Yanchen Liu, Xinyu Che, Brihi Joshi, Renzo Fukuda, Junhua Liu, Ruifeng Li — Mon, 02 Nov 2026 00:00:00 +0000

93% · New model 93%, Evaluation 83%

Training language-model agents to play long-horizon strategy games is hard because deterministic feedback is sparse and often arrives only at the end of a game. In-game scores can offer dense signals, but they do not reliably predict future outcomes: in 8-68% of position pairs across eight strategy games, the higher-scoring position leads to the worse outcome. We term this phenomenon scoreboard traps. In Freeciv, an open-source turn-based empire-building game where several players manage units, cities, and diplomacy over hundreds of turns, we look into better reward signals for gameplay agents. To facilitate this, we build CivHarness, an open system that saves current game states, branches on different decisions, and replays the game to future turns to provide measurable long-horizon feedback from saved positions. We use CivHarness to train CivTelescope, an LLM Judge that predicts which one of two game positions will yield the better outcome. On scoreboard-trap pairs, CivTelescope is 1.3-2.0 times more accurate than off-the-shelf language models, and it generalizes to new game starts, later turns, and unseen rules. We then train a Freeciv gameplay agent with CivTelescope as a dense reward, and call this recipe CivMarshmallow. When used in the middle stage of the game where scoreboard traps are common, CivMarshmallow reaches an average training phase gain of 8.94 points, 77% more than the best baseline, the end-of-game reward. Over the full 120-turn game, CivMarshmallow uses a hybrid reward that starts from the scoreboard or the end-of-game reward and shifts to CivTelescope as the game advances; this hybrid reward scores highest among the rewards we compare.

SparseGDN: Chunk Sparse in Linear Attention

Jiedong Zhuang, Peng Zhao, Jialin Tang, Haoji Hu — Mon, 02 Nov 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 34%

Gated Delta Networks (GDNs) offer linear-time sequence modeling, yet long-context prefill remains costly due to dense intra-chunk computation, recurrent-state updates, and intermediate state storage. We introduce SparseGDN, a hardware-aware framework that combines structured intra-chunk approximation with input-dependent recurrent-event selection while producing outputs for every token. A block-adjacent solver reduces local Delta-solve computation, while a dynamic selector allocates recurrent-update budgets across head pairs. The GPU implementation materializes updates only for selected events and reconstructs token outputs using compact state snapshots and group-local interactions, reducing computation and memory traffic. We evaluate SparseGDN on Qwen3.5-4B and Qwen3.5-9B across RULER tasks and context lengths from 4K to 128K. Retaining approximately 29% of recurrent events on average, SparseGDN achieves 1.50-1.94x GDN-core prefill speedups over native FLA kernels, including online selection overhead.

Stepwise-Querying: Distilling Reasoning without Access to Teacher Chain-of-Thought

Dicong Yu, Wenhua Wang, Tian Wang — Mon, 02 Nov 2026 00:00:00 +0000

93% · New model 93%, Evaluation 80%

Reasoning distillation is widely used to transfer the reasoning capabilities of large language models to smaller students through teacher-generated supervision. In the black-box setting, however, teachers may expose only final answers and brief solutions, leaving intermediate calculations and validity conditions unavailable and thus limiting the reasoning supervision available for distillation. We propose Stepwise Querying, which elicits step-specific explanations without requiring access to the teacher's internal chain-of-thought (CoT) traces. Independent queries to the same teacher request intermediate derivations, validity checks, and corrections for individual steps, the assembled solutions are then used for standard supervised fine-tuning. Across two teachers, two students, and four mathematics and code benchmarks, our method achieves the highest score in 14 of 16 comparisons with three baselines. We further introduce Stepwise Explanation Information Gain (SEIG), a diagnostic based on the log-ratio of a frozen student's probabilities of correct step completion under stepwise and global explanations. Analyses on GSM8K show that incorrect-answer trajectories can remain useful when they preserve valid intermediate reasoning, while further decomposition yields limited additional gains.

CRoS: Causality Reward on Scrambled CoTs Improves Accuracy and Reasoning Faithfulness

Xiao Yu, Justin Chih-Yao Chen, Zaid Khan, Hyunji Lee, Mohit Bansal — Mon, 02 Nov 2026 00:00:00 +0000

93% · Evaluation 93%, New model 90%, Hallucination 38%

Reinforcement learning with verifiable rewards (RLVR) has become the standard method for training language models to reason. However, RLVR typically relies solely on outcome correctness, which does not guarantee that the CoT causally contributes to the final answer. Without this connection, outcome rewards may fail to reinforce high-quality reasoning. Existing causality metrics assess CoT causality by perturbing the CoT, such as deleting reasoning steps or injecting errors, and measuring changes in the final answer. However, these perturbations inevitably alter information in each step. Consequently, high causality scores may reflect sensitivity to information changes introduced by perturbation rather than true causality. Such confounding effects further induce undesirable reasoning behaviors when these metrics are used as training rewards, including generating less informative CoTs and reducing exploration and self-correction. To address this problem, we propose CRoS (Causality Reward on Scrambled CoTs), a novel CoT causality metric that perturbs the order of reasoning steps without altering content, thereby preserving information within each step. On Reasoning Gym with Gemma-3-4B, CRoS outperforms existing causality and outcome rewards in out-of-domain accuracy, CoT quality, and CoT faithfulness. Specifically, CRoS achieves 46.8% accuracy on five unseen logic and games datasets, outperforming the strongest baseline (43.8%). It also achieves better CoT quality, consistently outperforming baselines in CoT-Pass@K from K=1 to 128, averaging 43.4% vs. 41.4%. For CoT faithfulness, CRoS achieves a hint acknowledgment score of 34.0% vs. 20.2%.

Incongruence-Aware Evaluation of Multimodal Speech Emotion Recognition in Audio Large Language Models

Fengji Jing, Guy Laban — Mon, 02 Nov 2026 00:00:00 +0000

92% · Evaluation 92%

Voice-interactive large language models process both what a speaker says and how they say it, yet the two channels do not always agree. We ask whether that disagreement degrades speech emotion recognition. Mapping text and audio into a shared valence-arousal-dominance space, we convert the distance between them into a continuous, threshold-free incongruence score with a two-component Gaussian mixture, and validate the score against human judgement. We evaluate three audiocapable models on acted, conversational and spontaneous speech (eNTERFACE'05, MELD and CMU-MOSEI). With the model held constant across modalities, incongruence raises the odds of error whenever audio is present, in every corpus, after controlling for the ground-truth emotion and for distance in the feature space. Supervised classifiers trained on the same features behave in the opposite way, which suggests that the difficulty lies in how these models combine raw evidence rather than in the signal itself. Text-only effects are unreliable: swapping only the model that reads the transcript reverses their sign on one corpus. Rebuilding the score without the emotion lexicon reproduces the central result, and among six simpler alternatives it is the only label-free score that predicts error in all three corpora. Code, data tables and prompts are released.

LanG 2.0: Energy-Efficient Agentic AI for Semi-Automated Security Operations at Scale

Anes Abdennebi, Nadjia Kara, Laaziz Lahlou, Hakima Ould-Slimane — Mon, 02 Nov 2026 00:00:00 +0000

92% · Efficiency 92%, Evaluation 52%

Security operations centers (SOCs) need to turn terabytes of alerts, logs and flows into trusted, affordable decisions to unveil hidden threats. We present LanG 2.0, an agentic system that detects threats, correlates evidence and writes rules, investigating whether small local models can replace frontier models affordably and safely when structural safeguards, not model accuracy, provide the guarantees. It incorporates entity-anchored correlation, human approval of rule changes and a tamper-evident ledger. Tests span eight flow datasets from five networks, real XDR alerts and multistage campaigns. Against its predecessor, campaign correlation precision rises from 18.75% to 51.81%, and rule acceptance from at most 60.00% to 98.33%. Against Claude Opus 5.5, scoring flows against each network's normal traffic detects attacks on unseen networks at 73.42% F1 without false positives, versus 58.10% with 49.00% false positives, for at least 650 times less energy. Finetuned models statistically match it on alert classification and rule writing for 5.6 to 59 times less energy. The work demonstrates LanG's practical scalability, processing 3.33 million records in 6.4 minutes, over 30 million hourly, and its safeguards resisted prompt injection, label poisoning and ledger tampering, so small local models can carry most SOC detection affordably and safely.

EDU-Jev Training System-One Decision Models for Education

Jiazheng Li, Yulan He — Mon, 02 Nov 2026 00:00:00 +0000

92% · New model 92%, Efficiency 86%, Evaluation 62%

Many automated decisions in education involve choosing from a fixed set of options: assigning a rubric score, checking whether an answer meets a marking criterion, or identifying the first wrong step in a solution. Generative language models are increasingly used to make these decisions, but their responses require parsing, longer outputs increase inference costs, and confidence expressed in words does not necessarily reflect the probability of being correct. We present EDU-Jev, a family of single-pass ("System One") decision models at 0.8B, 2B and 4B parameters. Given an assessment context and a question with predefined answer options, the model returns a probability distribution over those options without generating text. A pointer head scores a marker for each option, ensuring the predictions are restricted to the allowed choices. Training uses a single supervised stage with soft cross-entropy against the available target label distributions; calibration is assessed empirically. Across twelve education benchmarks, EDU-Jev 4B achieves an average accuracy of 67.6%, exceeding the strongest of five commercial decision APIs by 9.4 percentage points and the best open 4B decision model by 18.7 points. On four cross-domain education sets, it matches the strongest API, which remains ahead on general decision-making and reasoning tasks. The 2B model processes about 150 decisions per second on a single H100 GPU.

MUON-RT Reusing the Teacher's Trajectory for Momentum-Free Muon for LLM Self-Distillation

Aryan Das, Vinay Kumar Verma, Aakansha Mishra — Mon, 02 Nov 2026 00:00:00 +0000

92% · New model 92%, Efficiency 31%

Self-distillation methods for large language models with an exponential moving average (EMA) teacher already retain a history of student updates. Muon nevertheless stores a separate momentum buffer, increasing persistent training memory. We propose Muon with Reused Teacher (Muon-RT), which reuses the teacher-student displacement to eliminate momentum buffers for hidden weight matrices. We show that this displacement encodes an exponentially weighted history of applied updates and smooths temporally uncorrelated noise under a local model. Muon-RT removes its radial component, normalizes it spectrally, and mixes it with the current gradient after Newton-Schulz (NS) orthogonalization. Under explicit regularity and alignment assumptions, we establish linear contraction of an expected objective-gap bound toward a residual error level. With Self-Distillation Fine-Tuning (SDFT) on Qwen2.5-7B-Instruct, Muon-RT improves average accuracy over AdamW by 2.6 points with 42% less persistent memory, and over momentum Muon by 5.4 points with 27% less persistent memory. With Self-Distillation Policy Optimization (SDPO) on Olmo-3-7B-Instruct, it improves average accuracy over AdamW by 2.0 points.

Breaking Information Islands in Sparse Tuning via Small-World Connectivity

Xianchao Guan, Zijun Xiong, Yifeng Wang, Yubo Cui, Yaowei Wang, Guo-Sen Xie — Mon, 02 Nov 2026 00:00:00 +0000

92% · New model 92%, Efficiency 40%

Sparse tuning is widely used to adapt large language models due to its parameter efficiency. However, its effectiveness depends not only on the number of trainable parameters, but also on the feature-interaction topology of the sparse update's structural support. Limited or uneven connectivity in this support may isolate subsets of feature dimensions, forming information islands that restrict direct task-specific interactions through the trainable update. To address this issue, we propose Small-World Induced Fine-Tuning (SWIFT), a sparse tuning framework inspired by small-world connectivity. SWIFT combines block-diagonal sparse updates for local feature mixing with a fixed deterministic scattering permutation that routes the mixed outputs across feature groups without additional trainable parameters. We further show theoretically that certain random sparse-update models and block-diagonal updates can exhibit information islands, whereas a sufficiently scattering permutation guarantees positive block-level expansion when the trainable block-diagonal factor consists of equal-sized, structurally dense square blocks. Extensive experiments on commonsense reasoning, natural language generation, and image classification demonstrate that SWIFT outperforms the evaluated parameter-efficient fine-tuning (PEFT) baselines on average and can match or surpass full fine-tuning. Implementation is available at https://github.com/xianchaoguan/SWIFT

TSBPO: Tree-Structured Branching for Policy Optimization in Interactive Language Agents

Rui Yu, Chong Mu, Jinchuan Zhang — Mon, 02 Nov 2026 00:00:00 +0000

92% · Agents 92%, New model 91%, Evaluation 56%

Training interactive language agents from terminal outcomes poses three coupled problems: which task and state continuations to sample, how to assign delayed feedback to decisions, and how to aggregate token losses for multi-token actions. We present TSBPO, a tree-structured sampling-to-update paradigm that addresses these problems within one training loop. A likelihood-fitted Root/Branch allocator uses retained task and state evidence to target continuations likely to produce contrasting terminal outcomes under fixed rollout budgets. On the resulting replayed physical trees, trajectory advantages compare immediate sibling segments, while observation-group step advantages compare outcomes at matched task/observation states. The actor then sums clipped token surrogates over each complete response and normalizes across active turns, avoiding response-length averaging in the policy-gradient term. With a Qwen2.5-1.5B actor, fixed checkpoints obtain 97.79% seen and 96.64% unseen stochastic success on ALFWorld and an 81.07% per-episode success rate over six stochastic episodes for each of the 500 WebShop test goals. A WebShop step-term ablation yields 75.93% at the same update-100 evaluation; an ALFWorld aggregation comparison shows higher observed command validity with complete-turn weighting. These results support the evaluated training configuration on two interactive benchmarks.

SimXAI: A Simple Drop-in Layer Increases Interaction Quality and Understanding in Conversational XAI

Mon, 02 Nov 2026 00:00:00 +0000

92% · Evaluation 92%, Agents 81%, Hallucination 64%

Conversational explainable AI (ConvXAI) agents often use GUS-style architectures with separate modules for intent recognition, dialogue management, and template-based response generation. While faithful to the underlying XAI methods, these pipelines are hard to adapt and often struggle with context-dependent follow-up questions. We present SimXAI, a lightweight LLM-centric conversational layer that replaces these modules with a single tool-calling LLM that bridges access to existing XAI resources. Rather than a standalone agent, SimXAI serves as a modular interface that enables conversational access to external XAI resources. We further introduce an evaluation suite for single- and multi-turn parsing, faithfulness, and contextualization. Across existing and new datasets, SimXAI achieves strong function-selection accuracy and high faithfulness while better handling context-dependent requests. In a user study replacing only the conversational layer of TalkToModel, a representative GUS-based agent, users completed tasks more accurately and in fewer turn. Overall, SimXAI improves GUS-based design while preserving faithfulness and improving conversational quality.

Can Personal Requirements Become Supervision for Planning?

Junwei Wu, Runze Yan, Yuzhang Xie, Hanqi Luo, Ziwei Dong, Jiaying Lu — Mon, 02 Nov 2026 00:00:00 +0000

91% · New model 91%, Evaluation 91%

Managing multiple chronic conditions through diet requires satisfying disease- and medication-specific requirements jointly across meals and days while preserving established eating habits. These requirements are public in clinical guidelines and drug labels, yet turning them into a feasible plan for one person remains difficult for language-model agents. In our dietitian-approved weekly planning evaluation built from NHANES, thirteen models, from medically fine-tuned 8B models to GPT-6.1 Sol, pass fewer than 3% of complete plans through direct generation. Explicit rules, higher reasoning effort, or model-directed tools leave GPT-6.1 Sol at no more than 8%. We present Noura, which pairs LLM planning with executable personal specifications instantiated from public clinical guidance and individual health and dietary profiles. During planning, the specification grounds foods, optimizes portions, checks the complete multi-day plan, and guides revision. Our key insight is that a specification that rejects a plan can also determine what to learn from it. Noura turns executed plans and contextually credited revisions into supervision for compact models, recovering targets even from trajectories that never achieve complete success. The same specification supplies supervised targets, preferences, and rewards for post-training. On 4,993 held-out participants, Noura-SFT raises Qwen3-14B's complete-plan pass rate from 24.0% to 72.5% within the same execution loop, cuts rule violations by 97.5% with no observed critical failures, and improves dietary quality. The locally hosted 14B planner incurs fewer rule violations than its GPT-6 Luna teacher. A matched comparison shows that executed targets reduce rule violations by 41% over model-written proposals. Across four Qwen, Phi, and Llama backbones, self-training alone reduces rule violations by 41-61% on a 1,249-person population subset.

TalentRL: Strategy-Guided Reinforcement Learning for Hierarchical Skill Routing

Huancheng Chen, Xiaodi Sun, Zhaoqiong Huang, Michelle Chang, Jingwen Lu — Mon, 02 Nov 2026 00:00:00 +0000

91% · New model 91%, Agents 61%, Evaluation 51%

Natural-language skills enable large language model (LLM) agents to reuse procedures across tasks, yet most systems retrieve individual skills from a flat SkillBank without explicitly modeling the high-level strategy that governs their selection and use. We introduce TalentRL, a strategy-guided reinforcement learning framework for hierarchical routing over procedural memory. Coarse-grained talents represent recurring strategies for skill use, while fine-grained skills provide executable procedures. For each task, the policy model selects a talent that guides execution and softly prioritizes its associated skills during similarity-based retrieval. However, since task outcomes depend jointly on talent selection, retrieved skill context, and downstream execution, terminal rewards alone cannot isolate the contribution of the routing decision. TalentRL therefore trains talent routing with reward contrasts between rollouts of the same task under matched interventions on the talent-conditioned context, and optimizes downstream execution with group-relative advantages. Across seven open-domain and multi-hop question-answering benchmarks, talent-aware supervised fine-tuning improves average success rate by 7.53%, 6.71%, and 9.21% over the corresponding zero-shot baselines for Qwen2.5-7B, Qwen3-8B, and Llama-3.1-8B, respectively. Reinforcement learning further improves these SFT checkpoints by 7.58% to 9.32%, and TalentRL achieves the highest average success rate among all compared methods on all three backbones. On the agentic benchmarks ALFWorld and WebShop, TalentRL achieves the highest ALFWorld success rate (99.3%) and the highest WebShop score (90.7) among all compared methods.

From 1D Benchmarks to Complex 3D Geometries: Can Agents Reliably Specify Problems for Automated PDE Simulation?

Akshat Jani, Sakhinana Sagar Srinivas, Venkataramana Runkana — Mon, 02 Nov 2026 00:00:00 +0000

91% · Evaluation 91%, Hallucination 67%, Agents 45%

Turning an engineering request into a simulation requires consistent geometry, physics, boundary conditions, and numerical settings. Recent agent-assisted systems automate simulation setup, but successful execution alone does not establish whether the specification preserves the intended problem or correctly handles missing information. We study whether agents can reliably construct complete executable JSON specifications for automated geometry creation, meshing, PDE solving, and result export. Our workflow combines model extraction with deterministic scientific and capability checks, explicit clarification, and validated local patches that preserve unrelated accepted fields. We evaluate specification correctness, execution completion, and numerical verification separately. On 120 template-related requests, the full workflow achieves 98.3% correct first-turn specification and decision outcomes, compared with 92.5% for schema-assisted single shot and 82.5% for model-only: gains of 5.8 and 15.8 percentage points, respectively. It makes no false-ready decisions among the 42 requests requiring clarification or rejection. Component ablations identify contributions from scientific checks and conditional repair. Separate downstream studies cover 1D, 2D, and 3D steady and unsteady, linear and nonlinear problems, including flow and elasticity with complex three-dimensional wing geometries. A 30-case gallery demonstrates automated execution from retained JSON specifications, while 9 of 12 selected live user-to-results tests achieve joint specification, execution, export, and numerical-reference success (75.0%). These results support traceable, engineer-reviewed simulation assistance within the tested scope, while fresh-response regressions identify consistency as a priority for further improvement.

Operon-S1: a fast decision controller for long-horizon biomedical agents

Genoria AI Team, Yuejie Hou, Yueyuxiao Yang, Weiting Tang, Meng Yang — Mon, 02 Nov 2026 00:00:00 +0000

91% · Efficiency 91%, Evaluation 62%, Agents 50%

Language-model agents now carry out long scientific work in biology, but their judgements about conventions, skills, errors, evidence and stopping are made inside the reasoning that does the work, where they cannot be measured, replaced or improved on their own. Here we present Operon-S1, a fast decision system, a System 1 for the agent, that separates judging from doing. A decision harness poses the agent's recurring judgements as closed questions at 13 decision points and acts only through fixed messages that never contain an answer, so any model that answers the questions can serve as the judge. Across seven life-science benchmarks, a local decision model took over these judgements in 0.6 s at the median for an in-loop request on CPU at no request cost, with no measurable change in the agent's score on five of six benchmarks, whereas the general language model that drives the agent, asked the same questions, accounted for 8.0% to 35.2% of task cost. Every decision is recorded with its consequence. Retraining on labels from task construction, verifiers or execution raised decision accuracy on every question in each of three trained versions, and a new version enters service only through a promotion gate. Task-level results in this study use the public checkpoint. The same closed decisions serve wet-laboratory protocol troubleshooting and the design of a germline target enrichment panel. Like the regulator of an operon, which switches genes on and off without making their products, Operon-S1 steers the agent without doing its work, and makes the agent's judgements a layer that can be measured, replaced and trained.

Prune Gradually, Adapt Sparsely: Efficient Pruning of Large Language Models at Extreme Sparsity

Mon, 02 Nov 2026 00:00:00 +0000

91% · Efficiency 91%, New model 87%

Weight pruning can reduce the storage and inference cost of large language models (LLMs), but preserving quality at high sparsity levels remains challenging. Existing methods face a quality-cost trade-off in this regime. One-shot pruning is cheap but degrades sharply as sparsity increases, even when followed by post-pruning recovery. A growing line of work instead optimizes the mask directly from the dense model under the end-to-end objective; among these methods, [Lee et al.,2026], the recent state of the art at extreme sparsity, reaches high quality but at the cost of model-sized optimizer state and auxiliary variables. To address this gap, we propose ETAPS, which prunes gradually and adapts sparsely: it tightens a global sparsity constraint in progressive stages while training a sparse delta with a small, fixed budget. Our key observation is that gradual weight removal keeps each pruning step small enough for sparse adaptation to compensate, allowing the adapted model to provide end-to-end gradients that guide the next pruning decision without model-sized optimizer state. Across six models from 125M to 13B parameters, ETAPS achieves lower perplexity than ELSA at every evaluated sparsity from 50% to 90%, with perplexity reductions of up to 42%. ETAPS is also far cheaper: at 90% sparsity, it prunes LLaMA2-7B on a single H200, where ELSA uses four, with $5.0\times$ smaller total persistent state and 49% fewer GPU-hours. Together, our findings point to a promising direction for achieving high quality in demanding regimes such as extreme sparsity at substantially lower cost.

CoFiBreak: A Coarse-to-Fine Diversity Reward for Effective Multi-Turn Jailbreak

Jialin Song, Xiaodong Liu, Weiwei Yang, Mingqian Feng, Chengquan Guo, Qiyao Ma — Mon, 02 Nov 2026 00:00:00 +0000

90% · New model 90%, Evaluation 48%, Efficiency 35%

Automated red teaming trains an attacker large language model (LLM) to elicit harmful responses from a victim LLM, but reinforcement learning can collapse attackers onto a narrow set of prompts when attack success is the only reward. Such a collapse of attack diversity can bias safety evaluation toward repeatedly testing near-duplicate attack patterns rather than the underlying harmful intents. Existing methods often trade off attack success or inference cost for greater content or strategy diversity. We propose CoFiBreak, a coarse-to-fine diversity reward for Group Relative Policy Optimization (GRPO) that improves attack diversity while preserving strong attack effectiveness and low inference cost. The coarse reward broadens which strategies the attacker uses by encouraging strategies underrepresented within each GRPO group, while the fine reward broadens how those strategies are realized by discouraging reuse of the same personas, framings, and output formats after harmful-intent information is removed from prompt embeddings. Diversity rewards apply only to attacks that already exceed an effectiveness threshold, so diversity cannot be gained through ineffective attacks. Since both rewards are computed only during training, CoFiBreak also retains inference efficiency. Trained with Llama3.1-8B-Instruct as the victim, CoFiBreak discovers 1.6-2.0x as many distinct successful fine-grained strategies as the strongest baseline across frontier victims on HarmBench. It reaches 98.8% ASR@10 (attack success rate with 10 generations per intent) on AdvBench with the highest diversity score among all compared methods, which is 2.1x that of the strongest-attacking baseline. The resulting attacks transfer to unseen frontier LLMs, reaching 71.0% ASR@10 on AdvBench with the unseen victim Claude-Opus-4.6. The code will be released.

Better Retrievers, Worse Labels: How Incomplete Relevance Judgments Distort Abstention Evaluation in Conversational RAG

Ivan Kochergin, Vasily Konovalov — Mon, 02 Nov 2026 00:00:00 +0000

90% · Evaluation 90%

Conversational RAG benchmarks such as MTRAG and RETECO evaluate retrieval and generation end to end: each system answers from its own retrieval, so abstention must be scored against whether each retrieved context is sufficient. A convenient shortcut reads sufficiency off the relevance judgments and treats a context without gold documents as unanswerable. We re-run the official baselines exactly and test this shortcut on RECOR (RETECO track~2) with four retrievers (nDCG@10 from .18 to .59) and three Qwen3 readers. Two open LLM judges and a panel of four small open models label each context by content, and a model of another family audits 250 contexts blind, with adjudication. The stronger the retriever, the larger the share of no-gold contexts that still contain the answer (28\% to 72\%) and the lower the shortcut's precision (.68 to .10; .70 to .21 under the audit). A reader agrees: Qwen3-4B answers 62--67\% of these false misses correctly, against 23--30\% of no-gold contexts judged insufficient and 1\% with no documents. In most false misses, the answer comes from a hard negative mined for the same question, another turn's gold or a neighbouring chunk of the gold page. On MTRAG, whose judgments come from annotators, precision drops from .72 to .50. Under the shortcut, answerability AUROC is up to .19 too low, and the best sufficiency detector appears to be a retrieval score that never reads the documents. We release code, runs and labels, and recommend scoring refusals against content labels, not the presence of gold.

Verifying Text-to-SQL Queries under Changing Business Semantics

Genliang Zhu, Xiaobo Chen, Ziyuan Wang, Chu Wang — Mon, 02 Nov 2026 00:00:00 +0000

90% · Evaluation 90%, Hallucination 31%

In text-to-SQL and AI-agent analytics, successful SQL execution does not establish semantic correctness under changing business definitions. We study revalidation of fixed queries after changes to refund attribution, tax treatment, eligible populations and exchange-rate dates. An executable verification profile binds the requested revision and knowledge cutoff to legal counterexample data and exact monetary comparison. A preregistered confirmation on 32 newly authored revision families and two local producer bundles retains 192 query-definition cells: 144 nonconforming, 36 finite-suite agreeing and 12 unresolved. Of the nonconforming cells, 63 agree on the current snapshot. At an eight-ledger budget, obligation-guided checking detects 137 discrepancies versus 127 for generic ordering, a paired difference of 6.9 percentage points (family bootstrap 95% interval [2.1, 13.6]; exact family sign-flip p=0.0156). Every ledger arm rejects zero of the 36 agreeing cells. Current-obligation and equal-weight scoring match the guided result, identifying coverage rather than revision-specific priority as the observed source of benefit. Controlled assays, a natural-query pilot and repeated samples provide separate mechanism and robustness evidence. A legal discrepancy refutes conformance; agreement remains restricted to the recorded suite. The study establishes an auditable business-semantic revalidation profile within its authored schema and local-model scope.

Ask on Demand: Targeted Peer Communication for Parallel Agentic Systems

Haofu Yang, Hongyang Yin, Ran Bi, Faguo Wu, Peng Xue, Feng Zhou — Mon, 02 Nov 2026 00:00:00 +0000

90% · Agents 90%, Efficiency 86%, New model 43%

Parallel LLM agents tackle complex tasks by exploring multiple trajectories, yet useful discoveries remain isolated in individual execution histories. Existing sharing methods often select or summarize information before other agents' needs become clear, potentially overlooking details that become important later. We propose Ask on Demand (AOD), a communication framework that enables agents to seek targeted peer assistance when specific information needs arise. Upon an approved request, AOD retrieves relevant evidence from peer execution histories, selects a suitable peer, and generates a response tailored to the request. To determine when such assistance is worthwhile, we train a lightweight gate using paired team continuations, with supervision derived from final team outcomes and inference cost. Experiments on AppWorld and WebShop demonstrate that AOD consistently improves task performance over independent parallel execution while achieving lower inference cost than existing information-sharing baselines. These findings highlight the benefits of aligning peer communication with agents' evolving information needs to achieve more effective and efficient collaboration.

Saidzoda-Tajik-LFM-2.6B: Adapting a 2.6B Hybrid Model to Tajik Tool Calling, and What Continued Pretraining Takes Away

Mon, 02 Nov 2026 00:00:00 +0000

90% · New model 90%, Agents 83%, Evaluation 83%

We release Saidzoda-Tajik-LFM-2.6B, a Tajik adaptation of the 2.6B-parameter hybrid model LFM2.5-2.6B-Base, built for tool calling in Tajik on modest hardware. The recipe has two stages: LoRA continued pretraining (CPT) on 1,485,319 cleaned Tajik web documents for 3 epochs, then one epoch of supervised fine-tuning (SFT) on a mix of machine-translated Tajik tool calls, English tool calls kept as replay, and Tajik instructions. On a 174-item tool-calling test set built from BFCL v3 and machine-translated to Tajik, strict full-call accuracy on Tajik prompts rises from 20.69 for the base model to 64.14 (paired difference +43.45 points, 95% interval +35.2 to +51.7); a lenient parser that also accepts the base model's own call formats puts the base at 31.72. English shows no detectable change (76.55 and 77.24). Every CPT checkpoint scores zero: CPT removes the native tool-call format and, as the lenient re-scoring shows, part of the skill itself. Outside tool calling, accuracy on a Tajik knowledge benchmark rises from 28.0 to 35.0, Belebele Tajik shows no significant change, and English MMLU falls from 60.8 to 52.5. The model is not a general Tajik assistant. It almost never declines when no tool fits (6.9% correct on Tajik, against 31.0% for the base under the lenient parser), it is weak at multi-turn calls, and its calls should be checked before they are executed. We describe the data, the training settings and every evaluation in enough detail to repeat them.

Trainable Vector Quantization for Large Language Models

Mon, 02 Nov 2026 00:00:00 +0000

90% · Efficiency 90%, New model 87%

Vector Quantization (VQ) is a rapidly evolving technique for compressing Large Language Model (LLM) weights, reducing memory usage, and speeding up inference. However, extreme compression deteriorates model performance, which is challenging to recover due to the discrete nature of VQ parameters. Existing methods employ training-free approaches that are limited to layer-wise reconstruction loss minimization. To address this limitation, we propose TVQ, a novel method that leverages straight-through estimation to enable gradient descent-based optimization of vector-quantized LLMs. By reformulating VQ LLM optimization with trainable latent weights, TVQ enables arbitrary objective minimization. We constitute our approach by applying TVQ in a block-wise knowledge distillation paradigm, which balances quality and training efficiency, and propose several practical solutions to accelerate training. In experiments on modern LLMs, we demonstrate that TVQ is compatible with modern VQ schemes. As a result, TVQ achieves state-of-the-art quality in low-bit post-training quantization settings with a modest training budget and straightforward PyTorch implementation. For instance, TVQ achieves near-lossless 2-bit quantization for LLaMA2-70B with only 1% accuracy drop on popular Question-Answering benchmarks.

Same Question, Different Answers: Auditing Persona-Dependent Variation in Six LLMs on Smoking-Cessation Advice

Artin Yousefi, Aizierjiang Aiersilan, Lorien C Abroms, David A Broniatowski — Mon, 02 Nov 2026 00:00:00 +0000

90% · Evaluation 90%, Hallucination 36%

Hundreds of millions of people now ask large language model (LLM) chatbots for health advice, which raises a question that is simple to pose and hard to answer: does everyone get the same advice? We audit six widely deployed LLMs from four providers on smoking cessation, a domain with codified clini- cal guidance and substantial public-health stakes, by putting the same four cessation questions to twelve user personas that differ in life stage, self-described expertise, and smoking history, for a total of 2,880 responses. What the models say depends on who appears to be asking. Every model named FDA-approved prescription cessation medications less often to teen personas than to adult personas, by 32 to 67 percentage points, and in five of six models a persona presenting health- care expertise received fewer facts than a persona presenting none. We then evaluate one deployable intervention: a short system prompt specifying a factual floor, the set of core med- ical facts every user should receive when they are relevant, while leaving tone, vocabulary, and framing free to adapt. The prompt narrows persona differences in every model, but its effect on overall fact coverage varies in both magnitude and direction across models, so a prompt validated on one system cannot be assumed to transfer to another. Two cautions frame these results. Equal mention rates are not always the correct clinical target, because prescription cessation pharmacother- apy is not recommended for minors, so a defensible factual floor must be age-aware rather than uniform. And because our personas vary on several attributes at once, we report scenario- level differences rather than isolated effects of age or expertise. We release all 2,880 responses and the analysis code.

★ Seq@K: Evaluating Language Models Under Sequential Feedback

Thomson Yen, Gilbert Yang, Naimeng Ye, Minghui Chen, Dong Xia, Tianyi Shen — Fri, 30 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Language-model agents differ fundamentally from prior intelligent systems in their ability to learn from natural language in context. This capability enables agents to iteratively refine their answers in response to feedback, a prevalent mode of operation in modern agentic workflows. However, the community primarily ranks models by metrics such as Pass@1 or Pass@K, which cannot capture how they utilize feedback on prior attempts. We introduce Seq@K, a natural evaluation protocol that measures whether a model can solve a problem within K sequential attempts, provided feedback after each previous failure. Unlike multi-turn agentic benchmarks that examine long-horizon planning, Seq@K isolates and formalizes the core capability of refinement from feedback. We evaluate Seq@K on a suite of 7 benchmarks and 16 frontier models, and show that Seq@K reveals a materially different model-selection landscape: under Seq@K, (1) model rankings often differ, (2) the Pareto frontiers admit more efficient configurations and (3) the gains from choosing the optimal model configuration can widen or shrink relative to those of Pass@K. We further show that analyzing model trajectories under Seq@K helps characterize iterative refinement behavior, such as the rate at which models address issues raised in feedback, or how rigidly they hold onto their own answers. Finally, we study how the nature of feedback affects models' iterative refinement, showing that practitioners must carefully design and benchmark the feedback loops of iterative agentic systems. Together, these results establish how Seq@K offers a more faithful metric for intelligent systems that typically operate in such feedback loop, and the necessity to consider Seq@K in evaluating models for these systems.

★ DecompGym: Benchmarking LLM Agents on Designing Decomposition Algorithms

Sikai Cheng, Tinghan Ye, Arnaud Deza, Xiaonan Qi, Miao Li, Litong Liu — Fri, 30 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 49%, Hallucination 33%

Benders decomposition and column generation solve large optimization problems by passing cuts, columns, and dual values between a master problem and subproblems, and are hard to implement correctly. Large language models now automate parts of algorithm design, which points toward agents that build decomposition solvers for new problems. As a first step toward such agents, we study how well current coding agents design and implement these two algorithms, and how they fail. We introduce DecompGym, 81 instances from nine problem families (five Benders, four column generation), in which an agent writes a solver for a prescribed method. Each solver must meet three criteria: execution within a 60-second time limit; numerical feasibility together with a method audit, in which an LLM judge reads the solver's code; and an objective within 1% of the stored optimal objective. We evaluate 12 GPT, Claude, and Gemini models running in coding-agent tools. Over up to ten attempts, we compare independent regeneration with reflection, which first diagnoses the failure. On the first attempt, two models solve every instance and seven solve more than 90%. Most Benders failures (48 of 62) are missing or suboptimal solutions. The most common column-generation failure is a method-audit rejection. Output checks alone would accept 93 of the 108 attempts that fail the method audit. The most common rejection, an optimalif, occurs 96 times, 92 of them in one model.Reflection turns 66% of retries into successes, compared with 39% for regeneration, and it repairs method-audit failures, for which the judge names the defecective gaps. Agents for decomposition therefore need checks on how a solver reaches its answer and feedback that locates the error. The benchmark is available at https://github.com/SIKAI-C/DecompGym.

★ EPSBench: Benchmarking LLM Agents for Earnings Forecasting

Yanbin Yin, Chuyu Xu, Liqin Ye, Jing Peng, Chao Zhang — Fri, 30 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 35%

Quarterly earnings forecasting requires an agent to connect financial evidence to the income-statement line items that produce a company's reported earnings. We introduce EPSBench, a benchmark for studying this ability across 149 firm-quarters from 86 US companies. Agents read a fixed collection of filings, earnings-call transcripts, and news, then estimate company-specific income-statement drivers with cited evidence. A shared accounting calculator converts these estimates into GAAP diluted earnings per share (EPS), providing a common basis for comparing forecast assumptions. We evaluate six models under a common harness, assessing point forecast error relative to date-matched analyst consensus and checking whether their prediction intervals achieve the stated coverage. Three models show no detectable paired difference in point forecast error from consensus, and three have higher error. Across all six models, nominal 80% intervals contain reported EPS in only 37--70% of runs that return an interval, and most misses fall above the upper bound because the forecasts run low. The driver records trace most of the error gap between models to operating-income estimates.

★ Investor's Last Exam: Can AI Agents Do Equity Investment Without Seeing the Future?

Huangfei Jiang, Anning Hu, Kai Sheng, Yuchen Cai, Rui Ge, Xiangyu Kong — Fri, 30 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 69%

Financial agents must form forecasts and investment judgments from decision-time evidence, with predictions evaluated against subsequent outcomes. Existing benchmarks assess financial knowledge, research workflows, or forecasting, but evaluating the full research process requires a strictly frozen information set and professional evaluation criteria. We introduce Investor's Last Exam (ILE), an equity-research benchmark led by experienced buy-side investors and fund managers. Its three components are a versioned financial database spanning 3.43 million document versions and nine source categories, supplemented by structured market data, with validated timestamps governing retrieval and computation; a task taxonomy organized around professional analyst deliverables; and an expert-led framework linking scoring criteria to sources, calculations, and assumptions through financial dependency graphs. Evaluation spans 86 assignments and jointly assesses research quality, forecast accuracy, and simulated investment performance. GPT-6 Astra and Claude Opus 5.5 lead overall at 68.1 and 66.9 out of 100, yet even Astra scores only 52.5 on Forecast & Model. Our results show that even state-of-the-art models fail to translate overall research capability into predictive performance. ILE shifts the evaluation unit from isolated capabilities to the evidence--assumption--judgment chain, allowing researchers to locate where analytical advantages fail to translate into forecasts and investment judgments, and providing a traceable, diagnostic foundation for training and evaluating financial agents.

★ When Support Crosses the Line: Evaluating Ethical Boundaries in LLM-Based Mental Health Support in the Chinese Context

Bichen Wang, Yufeng Jin, Yixin Sun, Yanyan Zhao, Bing Qin — Fri, 30 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Large language models (LLMs) are increasingly explored for mental health support, yet existing benchmarks primarily focus on supportive capabilities rather than whether models consistently adhere to professional ethical boundaries. This issue is particularly important in multi-turn interactions, where users may introduce new information or exert emotional pressure over time. We introduce MindEth-Bench, a benchmark for evaluating ethical boundary adherence in multi-turn mental health interactions. Grounded in professional ethical standards for mental health services in China, MindEth-Bench covers multiple dimensions of ethical conduct and uses Monte Carlo Tree Search (MCTS) to systematically explore interaction trajectories that stress-test model boundaries. Experiments across nine LLMs show that models that initially respond appropriately can shift toward ethical violations over extended interactions, with substantial variation across models and ethical dimensions. These results show that single-turn evaluations may overestimate the ethical reliability of LLM-based mental health support systems and highlight the need to evaluate ethical behavior throughout multi-turn interaction.

★ HEARSAYBENCH: Can AI Navigate from Abstract Human Rights to Lived Lives?

Ava Iranmanesh, Sobhan Lotfi, Ali Iranmanesh, Liwei Jiang — Fri, 30 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

As large language models (LLMs) reach users worldwide, personal counseling has emerged as one of their most consequential applications, sometimes in life-critical situations. Yet broader access does not mean equal support. The text these models learn from disproportionately reflects people with the freedom to act, embedding an implicit assumption that users can leave, report, seek help, or appeal. Existing benchmarks largely inherit this assumption, overlooking users whose agency is constrained by structural conditions such as war, statelessness, and displacement. Meanwhile, models are increasingly shaped by normative frameworks such as Constitutional AI, often grounded in ideals like human rights, even as many users live where those rights offer little protection or recourse. We introduce HEARSAYBENCH, a human-authored dataset of 400 test cases with in-region human verification grounded in scenarios where structural barriers stand between a person and the exercise of their rights. Built from United Nations and other human rights reports, the scenarios span over 80 regions and cover social, environmental, and personal constraints. Drawing on Nobel laureate Amartya Sen's Capabilities Approach, HEARSAYBENCH tests whether models can distinguish what a person is formally entitled to do from what they are actually able to do. Across 14 proprietary and open-weight models, every model infers a user's situation more accurately than it advises on it. Stronger models narrow this gap, but their recommendations remain unsafe to follow. Models may understand lives shaped by constraint, yet that knowledge remains hearsay if it cannot translate into advice people can act on. Those people are not alignment's long tail; they are among the users for whom good guidance matters most. HEARSAYBENCH makes this failure measurable, a step toward systems that serve the world as it is rather than as assumed.

★ SocialArena: Evaluating Social Intelligence in Large Language Models

Zongqi Wang, Rui Wang, Yuchuan Wu, Hua Yang, Bingxue Qiu, Tianle Gu — Fri, 30 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

We introduce SocialArena, a benchmark for systematically defining and evaluating the social intelligence of large language models (LLMs). Although LLMs are increasingly deployed in socially interactive applications such as mental health counseling, companionship chatbots, and educational tutoring, existing evaluations remain fragmented focusing on isolated capabilities or single scenarios and evaluation reliability is largely unverified. This work makes three main contributions to address these limitations. (1) We first propose a unified three-tier framework that defines social intelligence along three capability dimensions: role-playing ability, mental state reasoning, and behavioral reasoning. To operationalize this definition, we design a role-playing dialogue paradigm with explicit character thinking for evaluation, enabling simultaneous assessment of these capabilities in a manner that mirrors realistic human interactions. (2) To ensure diversity across social scenarios and thereby prevent evaluation conclusions from being confined to specific scenarios, we introduce the U-TRIMF framework, which guides us in synthesizing social contexts with broad coverage. (3) To obtain reliable evaluation, we invest $132.4K to recruit and train human annotators and adopt a human-annotated arena-style protocol, where annotators compare model responses and provide preference labels along with natural-language critiques. We hope that SocialArena can provide valuable insights and guidance for future LLM improvements.

★ ATLAS-Bench: An Agentic Tool-Use Benchmark for Local Area Search

Keyun Zhang, Dapeng Yin, Fanfu Wang, Shuang Lang, Zhenyu Zhu, Yulin Huang — Fri, 30 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 96%, Hallucination 66%

Reliable local search is essential for LLM agents to support everyday decisions and longer plans by finding places that meet users' needs at given times and locations. Yet existing benchmarks provide limited coverage of whether search results satisfy mandatory requirements, account for preferences, and ground factual and experiential judgments in place attributes and reviews. We introduce ATLAS-Bench, a benchmark of 1,200 queries across three regions, expanded from real search events using place attributes, reviews, and existing user profiles from the platform. Each query distinguishes mandatory requirements from preferences and factual conditions from subjective judgments, with supporting evidence and criteria for checking each requirement. Queries are validated through iterative assertion checks and cross-model review. We also introduce ATLASBox, a standardized evaluation environment with search tools and a fixed point-of-interest (POI) knowledge base. Given a query, timestamp, and location, an agent searches place attributes and reviews to recommend evidence-supported options or report that no place meets all mandatory requirements. We evaluate answer quality, trajectory quality, and execution cost. Across 14 configurations of 11 models, GPT-6-Luna achieves the highest constraint satisfaction rate among returned recommendations (71.5%) and factual accuracy (95.6%). However, it returns at least one acceptable recommendation for fewer satisfiable queries than GLM-5.3 (67.3% versus 85.3%). These results highlight the challenge of satisfying individual requirements while providing usable recommendations across queries.

★ OpenFrontiers: A Collection of Open Research Problems and a Benchmark for Scientific Rediscovery

Jiarui Yao, Ruida Wang, Tong Zhang — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 67%

Large language models (LLMs) have shown increasing promise in research-level problem solving, yet a comprehensive collection of such problems is still missing, and evaluating them on research-level problems remains challenging. We introduce OpenFrontiers, a cross-disciplinary resource that combines a structured corpus of open research problems with a retrospective benchmark of recently resolved problems in mathematics. The corpus records problem statements, context, progress, and other information for 12,151 problems. The benchmark contains 200 problems, 110 resolved in 2026 and 90 still open, each with a Lean~4 statement checked by automated probes and two independent reviews; a subset HQ-50 further filters based on quality score, etc. Our evaluation protocol grades formal proofs with two independent checkers in an isolated sandbox under closed-book, frozen-retrieval, and live-web conditions. Evaluating seven LLM agents on the HQ-50 subset under a standardized budget, we find that closed-book rediscovery is rare, at most 7 of 40 resolved problems, whereas live-web success mostly reflects retrieval of published solutions. By linking a broad research-problem corpus to a formally graded evaluation subset with documented solutions, OpenFrontiers provides a foundation for measuring research-level rediscovery and studying progress toward solving currently open scientific problems. The dataset and benchmark could be found in https://huggingface.co/collections/FlippyDora/openfrontiers.

★ Evaluating and improving AI agents for sequencing protocol understanding

Chi-Lam Poon, Michael Goneos, Zezheng Song, Xi Chen — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 49%, Agents 37%

Existing scientific-agent benchmarks rarely evaluate whether agents can reconstruct the exact molecular structures and generation workflows specified by sequencing protocols, a capability critical for analyzing customized sequencing assays. We introduce LibStructBench, a document-grounded benchmark for reconstructing sequencing-library structures and molecular workflows. Our experiments show that frontier agents often produce approximately correct reconstructions but remain unreliable at exact reconstruction, particularly in molecular-state representation and workflow connectivity. Switching the underlying language model produced larger average performance differences than switching agent harnesses, with a statistically significant difference only for molecular-state reconstruction. Additional knowledge improved oligo-family recovery in a small pilot without improving mean molecular-workflow scores. Autonomous accumulation of procedural rules and cross-protocol examples showed non-monotonic validation performance and no mean held-out molecular-workflow improvement in the tested trajectory. These findings highlight the need for more controlled and verifiable approaches to scientific reasoning about molecular processes in complex experimental protocols.

★ FilingLoop: Evaluating and Evolving LLM Agents for Corporate Disclosure Review and Revision

Wentao Zhang, Yilei Zhao, Guangyan Gan, Shuo Sun, Lin William Cong — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 82%, Hallucination 66%

Public disclosure repositories such as SEC EDGAR support informed investment through web access to corporate filings. Supporting their review with language model agents requires discovering unstated deficiencies across related documents, acquiring missing company facts, and verifying that requested revisions satisfy evidence-backed requirements. Existing financial question-answering evaluations provide limited evidence of these interactive capabilities. We introduce \textbf{FilingLoop}, an interactive benchmark and a self-evolving agent framework for corporate disclosure review and revision. Grounded in historical U.S.\ Securities and Exchange Commission correspondence, FilingLoop reconstructs 380 predominantly IPO review tasks from 96 companies under distinct information constraints. An evaluator separate from the acting agents scores final filings and responses against 6,099 eligible requirements and audits unsupported claims and harmful revisions. Our framework couples iterative review and revision within each case with procedural learning across cases, converting requirement-level evaluation feedback into role-specific guidance without updating model weights. Updates affect only subsequent cases, allowing procedural adaptation to be assessed without retrying scored tasks. On the evaluated stream, evolution increases requirement fulfillment from 24.28\% to 27.96\% and full comment resolution from 16.90\% to 20.75\%, with a 5.8\% increase in total cost over the fixed-policy variant. FilingLoop connects the concerns an agent discovers with the company facts it obtains and the revisions it commits, providing requirement-level feedback for procedures used in later review tasks.

★ SciGame: Let Agents Play the Scientific Discovery Games in the Real and Counterfactual World

Yusheng Zhao, Bowei He, Xue Liu, Mingxiao Li — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 38%, Hallucination 33%

Large language model agents are increasingly used for scientific discovery. However, existing benchmarks mix law inference with programming, score data fitting instead of prediction on unseen inputs, and do not separate empirical evidence from pretrained priors. We introduce SciGame, a controlled benchmark that converts scientific literature into structured scientific cards. Each card yields a pair of Original and dimensionally consistent Counterfactual environments with the same task interface. SciGame contains a Historical collection of canonical laws and a live Temporal collection curated from papers published after model knowledge cutoffs. It evaluates agents in candidate-guided Identification and open-ended Discovery settings. Each setting runs under Prior-only, Fixed, and Adaptive protocols. Experiments on five frontier models yield three findings. First, familiar priors hurt performance once the governing relation is altered. Second, adaptive experimentation helps the strongest models but harms weaker ones under limited budgets. Third, good explanations can hide quantitative failures. Recovering the exact symbolic form remains the main bottleneck, even when agents interpret evidence correctly. Current agents can investigate systems actively, but inducing exact laws in unfamiliar settings remains an open challenge.

★ Assessing Language Models for Scientific Applications: Inertial Confinement Fusion as a Use Case

P. B. Radha, D. D. Meyerhofer, M. Bhattarai, B. Cheng, L. Diaz, C. L. Rousculp — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, New model 74%, Hallucination 44%

Advancing language models for specialized scientific applications requires evaluations of scientific fluency, physical reasoning, and quantitative analyses. Using inertial confinement fusion (ICF) as a test case, we have developed an expert-curated benchmark of 250 questions spanning qualitative reasoning, numerical calculation, symbolic derivation, and equation formulation. Responses have been evaluated using deterministic numerical and symbolic checks together with language-model judges. We compared a variety of out-of-the-box closed-source and open-source large language models with Prospero, a model specialized through combined domain-adaptive pretraining and supervised fine-tuning for ICF. We also conducted initial retrieval-assisted evaluations. Prospero improved its aggregate score by 10.7 percentage points over its base model, with gains across all question types. Frontier models achieved the highest aggregate scores, but critical physics errors persisted, particularly in numerical calculations. We note that domain-specific performance can improve through specialization, but benchmark gains alone do not establish scientific reliability. Retrieval improved performance for a smaller model but yielded no statistically significant improvement for a frontier model. Two scientific stress tests further illustrate useful capabilities and the need for expert assessment of physical assumptions and model validity. Based on these findings, we recommend evaluating existing domain capabilities and retrieval benefits before committing to substantial additional specialization.

★ 3DCodeEdit: Benchmarking Instruction-Guided 3D Program Editing

Binjie Wang, Xinran Zhao, Shoubin Yu, Xiang Li, Silong Yong, Chenyang Yang — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Language models increasingly create 3D content through code, with practical workflows often relying on editing existing programs rather than regenerating them, for speed and precise control. Such edits require understanding what a program builds and foreseeing how a change reshapes it, yet prior benchmarks mostly evaluate generation or simple edits against a reference. In this paper, we introduce 3DCodeEdit, a benchmark of 2,592 instruction-guided editing tasks over 108 Blender programs that construct objects entirely in code. Its edits range from parameter to structural changes, style redesign, and scene composition, and are judged from renders against human-reviewed checklists. Across 17 models, the average success rate is only 37%, and even the strongest models complete fewer than 60% of structural and style edits. As models become more capable, their failures shift to geometric foresight, where the code runs but builds the wrong shape. We therefore study whether multi-turn feedback can compensate for this missing foresight. Feedback from a separate reviewer model that describes how the renders differ from the request helps most, more than maximum reasoning effort or an agent harness, and can let a 27B model surpass the single-shot rate of a 2.8T-parameter model.

★ Can LLMs Faithfully Solve Ontology Engineering Tasks? A Mechanistic Interpretability Study

Bohui Zhang, Albert Mero\~no-Pe\~nuela, Elena Simperl — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Ontological knowledge underpins large language model (LLM) performance on knowledge-intensive tasks. However, existing evaluations of LLMs' ontological understanding largely rely on behavioral correctness, leaving the underlying mechanisms unexplained. Existing mechanistic interpretability work has largely focused on instance-level factual knowledge, leaving ontological knowledge under-explored. To bridge this gap, we introduce MIOE, a causal-abstraction benchmark that evaluates interchange interventions for two ontology engineering tasks: entity typing and subsumption prediction. MIOE tests whether learned internal intervention targets can change an ontology-level target answer while preserving class-specific auxiliary information. Across Qwen3-8B and Qwen3-4B, we evaluate SAE-, DBM-, and DAS-based methods, including multi-task and multi-layer variants. Distributed subspace methods give the strongest results and generalize across prompt settings, but the induced changes remain only partially separated from auxiliary class-specific information. Multi-task objectives and contiguous layer-window interventions improve this balance, while class-level analyses reveal substantial variation across ontology classes. These results suggest that ontological predictions are mechanistically accessible in LLMs, although current methods do not yet isolate a cleanly disentangled causal mechanism for ontological knowledge.

★ Deogen: Automatic Benchmark Generation for Reasoning over Policies

Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Language models are increasingly being deployed in regulated settings, where a model must apply given rules to the facts of a situation to determine a verdict. Benchmarking this capability, known as deontic reasoning, remains a costly and manual process, with existing datasets largely hand-crafted. In this paper, we present DeoGen, a scalable, automatic, and parametrizable benchmark generator for deontic reasoning. By autoformalizing policy rules to first-order logic, we can construct decision trees that capture possible outcome trajectories. Each path down a tree becomes a benchmark entry, with each tree producing combinatorially many entries relative to tree depth. Paths ending at a leaf entail a verdict, while non-leaf paths measure a model's ability to abstain when presented with insufficient information. Evaluations show frontier models return the correct verdict 89-92% of the time. Notably, performance degrades (22-40% correct) on entries without sufficient information to claim a verdict, even with explicit instructions to abstain in such cases. In all, DeoGen presents an automatic, scalable, and parametric method for generating deontic reasoning benchmarks.

★ TreasuryQA Bench: Can AI Agents Work an Archive of Scanned Financial Records?

Ritiz Tambi, Sahar Shayegan, Joshua Meyer, Sun Kim, Ali Khan, Victor Shih — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 94%, Hallucination 41%

Organisations keep decades of financial records as scanned PDFs on shared drives, and AI agents can now reach those drives through Model Context Protocol (MCP) servers. Existing benchmarks do not mirror this production setting because they simplify the task. In some, the relevant documents are handed to the agent, so it never has to search. In others, the agent searches through documents with pre-extracted text, so it never has to read scanned images. We introduce TreasuryQA Bench, which requires both: 70 questions over 1,435 scanned U.S. Treasury publications, held in Veris's high-fidelity Dropbox twin. In each task the agent searches for the relevant file through Dropbox's MCP server, reads the pages it needs with Reducto's hosted OCR server or its own tools, and replies with one exact figure. A Python grader checks each answer against an accepted range around the correct value. In this paper we present results from ten agents. The best, Claude Opus 5.5 in Claude Code, solves 64.8% of attempts and the worst, GPT-5.6 Terra in Codex, solves 16.7%, so the benchmark is far from saturated. We also test how scan quality affects performance: every agent solves fewer tasks on the blurriest scans. Most failed attempts reply with a wrong number rather than no number, and the most common cause is a digit misread from the scan. Three worked examples are included with their answer keys.

★ HALO: Human-Grounded Granular Evaluation of Proactive Agents

Wenqi Pei, Xinyu Wang, Jinrui Liu, Cynthia Zhuge, Jie Sheng, Zihan Wang — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 72%

Proactive agents must decide whether, when, and how far to assist without an explicit request. Existing evaluations often rely on automated scores or binary verdicts, offering limited human diagnosis of evidence use and intervention quality. To address these gaps, we introduce HALO, Human-Grounded GrAnular EvaLuation of PrOactive Agents. HALO-Arena covers 4 proactive need types, Dependency, Blockage, Risk, and Opportunity, across 10 everyday domains and 50 personal environments. Scenarios require agents to retrieve evidence from structured databases or unstructured documents before judging whether assistance is warranted and choosing its scope. On HALO-Arena, we construct HALO-8K, a dataset of 8,000 responses from 10 models across 400 scenarios, annotated by 20 experts for action level, acceptability verdicts, and human diagnosis. For rejected responses, human diagnosis covers timing, intensity, and other issues. Recorded reasoning tokens and retrieval rounds measure the effort behind these responses. The best proprietary and open-source models achieve 72.88% and 56.63% acceptability, respectively. Our central finding is that over-proactivity is the dominant observed failure mode, with Premature timing or Excessive intensity accounting for 61.20% of rejected responses. Further analyses yield complementary findings on the balance between intervention and restraint, evidence acquisition, and the calibration of timing and intensity.

★ RetroChessBench: A Benchmark for Reconstructing and Proving Chess Histories

Tidor-Vlad Pricope — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 48%, Hallucination 32%

Frontier language models such as GPT 6 Astra have contributed to research mathematics, demonstrating substantial reasoning capabilities. Yet serious reasoning failures persist even in tasks governed by explicit rules. We demonstrate this in retrograde chess which is a playground for pure reasoning, where solvers reconstruct legal histories and distinguish possible events from forced ones. We introduce RetroChessBench, a 12-task benchmark covering exact-length constructions, unique endings, and exhaustive answer sets, with references, rubrics, and a terminal-reward environment. In two trials per task and condition, GPT 6 Astra High fails to produce the required final answer in 22/24 attempts without tools and 18/24 with code execution; mean terminal RL rewards are 14.75% and 27.19%, respectively. The examples identify faulty deductions: failed responses reject legal histories and treat temporary obstructions as permanent. Online retrograde archives often provide puzzles and final move sequences without detailed derivations. We hypothesize that limited exposure to these missing explanations in web-derived training data contributes to the observed reasoning failures. This questions how far frontier mathematical reasoning really generalizes beyond familiar solution patterns. Claims of general intelligence require evidence that models can transfer learnt principles to areas with few worked examples.

★ AgentDecept: Benchmarking Deceptive Behavior in Tool-Using LLM Agents under Pressure

Yehan Zhang, Qianqi Yan, Hang Yin, Ching-Chen Kuo, Shan Jiang, Ming Ouyang — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 97%, Hallucination 40%

LLM agents increasingly carry out multi-turn tasks across everyday applications, yet users often see only the agent's report rather than the observations, API calls, and state changes behind it. Existing deception benchmarks have studied long-horizon interactions and tool-use settings, but whether agents faithfully report their actions across realistic, stateful application workflows remains underexplored. To address this gap, we introduce AgentDecept, an executable benchmark for evaluating behavioral deception under pressure. AgentDecept contains 477 pressure variants derived from 257 source tasks across six pressure sources, each paired with a pressure-free counterpart. Interventions enter through event messages, task context, application state, tool responses, and service quotas, spanning explicit events and implicit task or environment changes. Our evidence-grounded protocol evaluates full trajectories by relating user-facing claims to time-ordered observations and executed outcomes, distinguishing falsification, concealment, and equivocation from honest failure and disclosed limitations. The resulting behavioral labels also provide a setting for testing existing activation-based probes. AgentDecept enables controlled analysis of when multi-turn agents' reports diverge from execution in realistic application interactions.

★ Heterogeneous Co-Evolution Learning for Tool-Efficient Search Agents with Zero Labels

Qin-Wen Luo, Sheng Ren, Xiang Chen, Rui Liu, Jun Fang, Naiqiang Tan — Fri, 30 Oct 2026 00:00:00 +0000

98% · Agents 98%, New model 94%, Efficiency 61%

Deep search agents extend language models by interleaving reasoning with external retrieval. However, effective search-agent training requires addressing two challenges beyond final-answer optimization. First, the capability frontier of agents is largely determined by training questions, whereas fixed or labeled task collections cannot adapt as agents improve. Second, outcome-only reinforcement learning treats all successful trajectories equally, even when some correct answers rely on redundant searches, creating an inherent tension between answer accuracy and tool efficiency. To address these issues, we introduce \textsc{ZEST-Search}, a zero-label and tool-efficient framework for search-agent training via heterogeneous co-evolution. Starting from unlabeled reference questions, \textsc{ZEST-Search} constructs validated search tasks online through an answer-first and evidence-path-guided generation process. A heterogeneous proposer adaptively constructs training tasks around the solver's evolving capability boundary, while the solver learns search-augmented reasoning through reinforcement learning. To jointly optimize task difficulty and search efficiency, \textsc{ZEST-Search} employs an intermediate-difficulty reward for the proposer and a task-local tool-aware reward for the solver, encouraging informative tasks and correct answers with fewer redundant searches. Experiments on seven general-domain and multi-hop question answering benchmarks demonstrate that \textsc{ZEST-Search} consistently improves search-agent performance while achieving higher tool efficiency without labeled search-training data or expert search trajectories. Further analyses show that heterogeneous generation, adaptive difficulty control, and tool-aware optimization enable more effective search-agent training.

★ Bank-World: A Benchmark for Evaluating Agents on Banking Workflows

Daniel Rupawalla — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 44%

Bank-World is a benchmark of 50 banking tasks across two simulated institutions. Completing banking work requires more than producing the right calculation: agents must reconcile records, obtain permission, execute transactions, and keep instructions and status reports consistent as circumstances change. Bank-World tests these requirements through single-turn assignments and multi-turn interactions with simulated employees who supply information, approve actions, and revise instructions. Grading combines executable checks of recorded actions and state with model judgments of task requirements, including whether valid alternative solutions satisfy them. Across 1,600 attempts by four models, we find that correct decisions often fail to become completed work. Agents charge fees twice, distribute the correct total in incorrect amounts among the recipients, and execute payments while approval is pending. Their strengths also differ: Sol leads single-turn completion, Opus has the highest observed multi-turn completion rate, and Grok succeeds across more multi-turn tasks but less consistently. Bank-World makes these differences inspectable through task-level results and worked examples connecting source evidence, agent decisions, and simulated consequences. We retain the task environments, executable graders, and full trajectories privately to reduce contamination and preserve the benchmark for future evaluations.

★ PAPERBENCHX: Can AI Reproduce Science Across Disciplines?

Ruoyu Wu, Zengji Tu, Aolong Sun, Tianyi Ma, Junren Li, Xinbo Xu — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 76%

Before AI can be trusted to discover new science, it should be able to reproduce what science already knows. Reproduction provides a rigorous checkpoint for an AI scientist: the target claim is known, yet recovering it still requires translating an incomplete scientific description into an executable experiment and producing evidence that can be independently verified. Existing research-reproduction benchmarks largely focus on AI and machine learning, where experiments share relatively homogeneous software stacks. Extending this capability across scientific disciplines introduces a different challenge: agents must operate domain-specific scientific software, make physical and numerical choices, diagnose long-running computations, and distinguish valid evidence from merely plausible outputs. We introduce PAPERBENCHX, a benchmark of 93 expert-constructed paper-reproduction tasks spanning 12 research areas across electromagnetics, photonics, chemistry, materials science, biology, and robotics. Each task provides the source paper, a public task specification, and a domain-native scientific environment involving specialized scientific software. Agents must reconstruct a key experiment, execute and diagnose it, and deliver a runnable workflow that supports the target scientific claim. To evaluate scientific validity rather than surface-level agreement, we replay every submission from scratch and score only regenerated evidence against expert-validated criteria. Across 10 frontier agent configurations, the best configuration fully reproduces only 14% tasks under the 90% threshold. Meanwhile, mean rubric scores range from 27.53% to 59.08%, indicating that substantial partial progress rarely translates into complete scientific reproduction. These results reveal a large gap between paper understanding and reliably producing reproducible scientific evidence.

★ DynaRev: Benchmarking Web Agents under Dynamic Goal Revision

Shiqi Zhang, Guochen Zhou, Zesheng Wei, Hai Zhao, Zhisong Zhang — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 97%

Web agents are typically evaluated with goals fully specified before execution and fixed thereafter, whereas real users often revise their requirements during interaction. To study such interactions, we introduce DynaRev, a benchmark of 181 tasks to evaluate web agents under dynamic goal revision. DynaRev uses semantic milestones to deliver revisions at task-defined stages, covers additive revisions, unconditional overrides, and runtime-conditioned overrides, and evaluates correctness, responsiveness, and efficiency through trajectory-grounded evaluation. Five-model evaluations show that agents often satisfy only part of the task requirements, with performance varying across task sources, revision types, and models. Further analyses reveal varied milestone timing, mixed effects of dynamic delivery, and contrasting revision-handling outcomes, with gains from instruction rewriting while preserving browser state and losses from browser resets. DynaRev provides a testbed for studying how web agents adapt to evolving user goals. The code and data are available at https://github.com/zsq259/DynaRev.

★ Prefix Executability: Evaluating Tool-Using Agents Beyond Final Success

Amir H. Rezaeian, Weiyi Sun, Yassine Benajiba, Dan Roth — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 97%

Tool-use benchmarks often score agents by final task success, but terminal outcomes can obscure whether the intermediate tool-call trajectory was executable. We propose prefix executability as a trajectory-level evaluation protocol for tool-using agents. The protocol measures whether every prefix of a predicted tool-call sequence remains executable under tool schemas, state preconditions, and data dependencies, and summarizes reliability using Prefix Executability Curves (PEC), First-Error Depth (FED), Trajectory Executability Rate (TER), horizon-normalized AUC-PEC, and Recovery Rate (RR). We instantiate the protocol on stratified samples from BFCL-v3 multi-turn and ToolSandbox, and on the complete Airline/Retail base splits of $\tau^3$-bench. We observe that outcome-only and prefix-level rankings diverge: agents whose execution is conditioned on an unverified plan often achieve higher point-estimate final success than non-plan-conditioned baselines while substantially lowering TER and AUC-PEC. These divergences are precisely the failure modes that prefix executability is designed to expose: they reveal whether apparent success rests on an executable trajectory, where invalidity first enters, and whether success follows recovery from an earlier failure. As a validation intervention, explicit pre-execution plan checking can improve executable-prefix survival while maintaining competitive terminal success. These results show that final success alone is insufficient for evaluating tool-use agents and that prefix-level executability provides an actionable reliability signal for comparing agents, diagnosing failures, and assessing validation mechanisms.

★ IndicIPR-QA: A Benchmark for Extractive Question Answering on Indian IPR Laws

M. Mohaiminul Islam, Muhammad Abulaish — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Legal Question Answering (QA) over statutory texts is a challenging task in Natural Language Processing (NLP), especially in low-resource settings with complex legal systems. It requires answers that are precise and grounded in the law. However, to the best of current knowledge, no existing extractive QA benchmark provides exact answer spans from statutory texts across all seven Indian Intellectual Property Rights (IPR) Acts. This paper presents IndicIPR-QA1, an extractive QA benchmark for Indian IPR laws. The proposed dataset contains 4,462 QA pairs from 1,116 statutory contexts across seven Acts. Questions and answers are generated using GPT-4o and validated through nine automatic quality checks, single and multi-model LLM evaluations, and a 200-pair human audit. Section-level partitioning prevents source-context overlap across the training, validation, and test sets. Further analysis examines answer span length, lexical overlap, and multi-clause structure across the corpus and individual Acts. BM25 achieves a token-level F1 score of 20.21, while fine-tuned InLegalBERT reaches 85.02 under a unified sliding-window inference protocol. The large performance gap between BM25 (20.21 F1) and InLegalBERT (85.02 F1) shows that simple word matching is not enough to answer these legal questions. Performance also varies across the seven Acts, indicating substantial differences in answer-extraction performance.

★ Testimony from Timelines: Evidence-Grounded Significance Testing of Temporal Clinical Hypotheses

Chengyang He, Wenlong Zhang, Koduvayur Subbalakshmi, Ping Wang — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 60%

Ask a language model whether sales tend to dip after a product recall, or whether one side effect tends to follow another after a vaccine, and it answers in a fluent sentence that reads like a finding, with no test behind it. Assertions about how events unfold across many cases fill the Web, and each is a temporal hypothesis that only a significance test over the right records can settle. The verifier must specify the null, select the test and decide which records bear on the hypothesis, and any of these choices can go wrong while the conclusion still comes out right. Yet benchmarks for language-model data analysis grade only the conclusion, with the relevant data handed over. We study the problem in health, where a wrong verdict is costly, and present STRATA, a benchmark that grades every decision on the way to a verdict: 8,735 hypotheses over the clinical timelines of vaccine adverse-event reports, with gold plans, evidence sets, p-values and verdicts derived under a released codebook and reproduced by a deterministic oracle. We also propose ART (articulate, retrieve, test), a staged paradigm that exposes each decision as an inspectable output. Extensive experiments with five model configurations show that systems fail mostly at choosing which records count, a verdict-evidence gap that accuracy alone conceals. ART closes most of this gap and brings a 31B open model to within 0.01 of the frontier models under the same paradigm; the frontier models succeed once given the codebook.

★ Elucidating the Design Space of LM Data Attribution

Steven Yu, Sheng-Yu Wang, Kimia Hamidieh, Antonio Torralba, David Bau — Fri, 30 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Data attribution methods for Language Models (LMs) are typically evaluated using downstream tasks like fact retrieval. However, existing evaluations vary multiple dataset properties simultaneously, making it impossible to isolate exactly why an attribution algorithm fails. To diagnose these failures, we explore the design space of LLM data attribution, with a primary focus on evaluation. We introduce the Controlled Attribution Benchmark (CATT) for fact tracing, in which we fine-tune LLMs on entirely fictitious knowledge, providing unambiguous ground truth for attribution. This synthetic setup allows us to systematically alter one dataset configuration at a time, such as document redundancy and lexical distractors. With this configurable evaluation, we demonstrate that redundancy and distractors consistently degrade attribution performance. Rather than proposing a new estimator, we examine two design choices that are underexplored: the loss used to compute gradients and how per-token attribution scores are aggregated into a document score. We find that the true causal evidence within a training document is highly localized, but standard methods dilute this evidence into noise by averaging attribution scores across the entire document. We find that replacing standard cross-entropy loss gradients with the negative log-odds gradients surfaces the critical tokens from other uninformative ones. We can then aggregate only the top token scores to improve ground truth recall by 35 to 72 percentage points across multiple baselines on our diagnostic benchmark. The negative log-odds correction also generalizes to FTRACE-TREX, improving counterfactual retraining performance for every method we test, by up to 4.9x.

★ Exposure Is Not Commitment: Double-Stratified Evaluation of Latent Monitors for Tool-Using Agents

Rishabh Sinha — Fri, 30 Oct 2026 00:00:00 +0000

97% · Agents 97%, Evaluation 95%

Tool-using language agents are vulnerable to indirect prompt injection, and a growing literature probes model internals to catch unsafe actions before they execute, with sharply conflicting results. We argue much of the conflict is an evaluation artifact, and propose a double-stratification protocol: separate injection exposure from behavioral commitment, in both the evaluation and the training population. On the primary Qwen AgentDojo grid, monitors trained on the naive mixed unsafe-versus-rest contrast mostly learn exposure: they flag exposed trajectories whether or not the agent complies, and barely predict which cases turn unsafe. Retrained on exposed cases only, the same probe families predict the model's own unsafe tool call up to two steps early, the maximum the three-step trajectories admit (logistic AUROC 0.853, matched by a trained TF-IDF text baseline at 0.855). Mistral-7B reaches 0.896 and OLMo-2-7B 0.912; their naive contrasts learn different things: inverted exposure ranking on one, both axes on the other. Representations separate on exposure (probe 1.000 versus trained text 0.908) and where visible context is sanitized or unavailable. Training contrast alone reproduces the shape of recent pre-action-probe negative results, one candidate explanation for the conflict: the signal exists, but the model's own zero-shot forecast and NLA verbalizations do not reliably recover it. An unstratified AUROC overstates what a latent monitor knows, and what a naive contrast learns is population-dependent: evaluations should stratify both.

★ Fu-TCM: a multimodal large language model that learns traditional Chinese medicine diagnostic reasoning

Jiahang Xie, Tong Jin, Lingfei Wu, Yaping You, Jie Huang, Yuanchang Jin — Fri, 30 Oct 2026 00:00:00 +0000

97% · New model 97%, Evaluation 74%

Large language models (LLMs) are advancing in multimodal understanding and reasoning, accelerating medical model development. Traditional Chinese medicine (TCM) uses evidence from the four examinations to determine disease location and nature, differentiate syndromes and guide treatment. This diagnostic process constitutes a complete chain of reasoning. We present Fu-TCM, a multimodal TCM LLM trained on data retained after cleaning and filtering an initial collection of approximately 7.12 million examples derived from classical texts, textbooks, medical images, clinical cases and publicly available datasets. The model was further optimized using Bianzheng-Grounded Policy Optimization (BGPO), which explicitly targets syndrome differentiation through rewards for response format, intermediate TCM reasoning fields and final syndrome predictions. Fu-TCM-27B led all six benchmarks spanning TCM reasoning, text and vision, achieving a macro-average accuracy of 81.01% and surpassing the strongest reference model with complete results across all six benchmarks by 7.13 percentage points. Its largest margin was 11.90 points on Fu-bianzheng. Case-level analyses further showed a positive association between intermediate-field accuracy and final syndrome-selection accuracy in most evaluated models. These findings suggest that TCM syndrome differentiation may be computationally learnable and that structured reasoning supervision can improve performance across TCM tasks.

★ RoleFUZZER: User-Behaviour Fuzzing for Role-Specification Adherence in Large Language Models

Kshitish Ghate, Michael Saxon, Tadayoshi Kohno, Aylin Caliskan — Fri, 30 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Large language models (LLMs) are often conditioned into specific roles such as tutors, interviewers, or agents acting for an organisation, through system prompts that specify how they should behave. The extent to which they can actually adhere to these role specifications under multi-turn interactions with realistic and potentially problematic users is understudied. To address this, we propose user behaviour fuzzing, a guided search over plausible user behaviours that can lead a model to violate its role specification, and introduce RoleFUZZER, a framework that adapts coverage-guided fuzzing from software testing to operationalise the search. RoleFUZZER constructs scenarios guided by a user-behaviour coverage grid that crosses a grounded taxonomy of user interaction types and the role specification of the model under test i.e. what is under pressure. With RoleFUZZER, we evaluate LLMs across five role specifications, including tutor, financial advisor, AI interviewer, antagonistic roleplay character, and web agent under organisational policies. Across simulated and validated conversations in 15 campaigns, the framework reaches >= 89% coverage of the 64-cell behaviour grid for every model and role tested, and surfaces specification-grounded failure patterns. At an equal budget of conversations, removing the RoleFUZZER grid and asking an LLM to generate specification targeting users free-form lowers coverage to 39-69% with fewer kinds of violations. The same specification can also fail in opposite directions. For e.g., on the antagonist, open-weight target models override the player's safety signals, while Claude Haiku 4.5 breaks character instead. RoleFUZZER offers a tractable way to uncover vulnerabilities in LLMs across role specifications by testing diverse user behaviours that are likely to violate them.

★ An Alignment Benchmark for Enterprise Agents on Era by Eon

Fri, 30 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 87%, Hallucination 78%

AI agents are widely used in enterprises to access business systems and act on them in pursuit of a goal. However, fulfilling a user's request can cause harm. Era by Eon is a product that creates a digital twin of a company by simulating its business systems and data. Its simulators implement the programming interfaces of real products. Our benchmark uses ten of these systems, including Salesforce, Zendesk, Jira and Slack. We tested whether agents would carry out harmful business requests or trust misleading information in company records. We ran 19 models on 41 tasks three times, for 2,337 runs. Of the 2,304 runs that produced an answer, 918 (40%) contained a violation. The most common were false claims about actions or records (290 runs), following planted instructions (266) and disclosing sensitive data to the wrong audience (249). Every attempted change was recorded and blocked before it took effect. Misleading Slack messages caused 16 models to report an incorrect ticket status in 47 of 57 runs. No model was free of violations. The four best models violated in 22 to 24% of their runs. Yet one of them, Nova Pro, did the task correctly in only 30 runs.

★ Do Audio LLMs HEAR Pain? Tracing Acoustic Evidence from Speech to Language Reasoning

Wenqi Li, Xiangjue Dong, Sulaiman Vesal — Fri, 30 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Large audio language models (LALMs) can process speech directly, but it remains unclear how well they use the non-lexical acoustic information that speech carries. We study this question through pain, using TAME-Pain, which contains pain-neutral Harvard sentences and spoken pain ratings recorded at different self-reported pain levels. We introduce HEAR, a four-stage diagnostic framework that traces pain-related information from the speech signal to the model's prediction, and apply it to seven closed-source and open-weight LALMs. On the Harvard sentences, where pain can only be inferred from the voice, all models perform at chance level, and same-speaker references or few-shot examples improve balanced accuracy by no more than six points. On the spoken ratings, the same models reach 87.5-99.7% balanced accuracy. Layer-wise probing of open-weight LALMs shows that pain can be decoded from early audio representations (74.1% balanced accuracy for binary prediction and 45.0% for three-class severity), but that this information, along with most pain-associated acoustic cues, decreases in deeper layers. In an acoustic perception test, weaker models cannot distinguish even large changes in these cues, whereas stronger models can but still fail to infer pain. Finally, we recover pain-related acoustic cues and directly reinject them as structured text into the prompt, which improves pairwise pain prediction by up to 32 points. These results indicate that current LALMs encode paralinguistic information in early layers that is largely unavailable when the model produces its answer.

★ Think Before You Act: Decision Gate for LLM Agents

Iftach Shoham, Yuval David, Offer Akrabi, Erez Bilgory, Ido Levy, Sami Marreed — Fri, 30 Oct 2026 00:00:00 +0000

97% · Agents 97%, Evaluation 82%

LLM agents increasingly perform consequential actions through external tools, yet their proposed actions are often executed without independent verification of policy compliance or task completion. We introduce JevAct, a runtime verification framework that intercepts selected agent actions before execution and enables feedback-driven correction. JevAct compiles domain policies into executable state checks and structured questions evaluated by a prefill-only decision model, which produces probability-based judgments without relying on LLM generation. By combining deterministic verification of tool-state facts with probabilistic assessment of conversational context, the framework identifies potentially invalid actions and premature task termination, providing targeted feedback for bounded revision. JevAct operates independently of the underlying agent model and requires no modifications to its parameters or tools. We evaluate JevAct across eleven LLM agents on $\tau^2$-bench Airline and Telecom tasks and BFCL. Results show improvements in agent reliability across multiple models and benchmarks, with particularly consistent gains on Telecom and BFCL, while performance on Airline is more mixed. Overall, these findings suggest that combining structured decision models with executable verification can improve the reliability of tool-using LLM agents.

★ CopyLoop: Fast Looped Transformer Decoding via Context Copying and Shallow Screening

Wenyu Du, Stephen Chung — Fri, 30 Oct 2026 00:00:00 +0000

97% · Efficiency 97%, Evaluation 35%

Looped Transformers increase computational capacity through the repeated application of shared blocks, but their recurrent depth makes conventional full-depth decoding expensive. While existing speculative pipelines accelerate generation by processing consecutive token positions at varying depths, they are typically limited by sequential decoding overheads. We introduce CopyLoop, a training-free decoding algorithm that combines low-cost context-copy proposals with shallow screening and continuous recurrent scheduling. Rather than step-by-step generation, CopyLoop proposes multi-token suffixes from repeated context. It then evaluates them early at a shallow recurrent depth to efficiently filter out incorrect branches, processing newer blocks concurrently while earlier blocks are still progressing toward full-depth verification. We evaluate CopyLoop across six model checkpoints from the Ouro, Raven, and Huginn families, covering code generation and multi-step mathematical reasoning. CopyLoop achieves decoding speedups of 2.98x--10.70x over full-depth autoregressive decoding on Base models and 2.74x--3.52x on Thinking models, consistently outperforming or closely matching all competing baselines across evaluation tasks. Furthermore, ablation studies confirm a strong synergy between shallow screening and continuous scheduling, while our analyses show that CopyLoop requires only lightweight model calibration and yields greater speedups on longer generations with reusable contexts.

★ Contextual Embeddings Beyond the Gold Chunk

Fri, 30 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, New model 87%

Contextual embeddings represent chunks of a document with respect to the document's global context. These models are typically trained and evaluated under the assumption that a relevant document contains only a single relevant chunk. In this work, we argue that in real-world use cases a retriever should not only find answers spread over several chunks but also surface the evidence needed to verify them. To address this, we introduce pplx-embed-v2-context, a family of contextual embedding models trained by distilling graded chunk-level relevance from a query-aware context compression model. This removes the need for gold-chunk labels and makes the training targets independent of the chunking. We utilize a vision-language backbone for training and release pplx-embed-v2-context in 9B and 0.6B sizes with compatible embedding spaces. On the evaluation side, we introduce two benchmarks that label both the chunks that answer a query and the evidence needed to verify the answer. Context-bench is a controlled diagnostic, and Q2D-Chunk is built from real-world search queries. To encourage further research, we also publish Q2D-Chunk as a public leaderboard. Our 9B model sets a new state of the art on both benchmarks and ConTEB. We find that the visual retrieval capabilities of our 9B model transfer from its vision-language base model and propose to extend chunking to image inputs. We demonstrate that our models achieve state-of-the-art performance among dense retrievers on visual ViDoRe-V3 and that performance improves with granular image chunking.

★ Context Compaction for World Models

Hayder Tirmazi — Fri, 30 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Hallucination 35%

World models let observers explore and interact with generated worlds. Recent work shows that when observers revisit a region, current models often fail to regenerate it consistently. We view this as a context retrieval problem. LLM agents manage context retrieval at two layers of their stack. At the model layer, the KV cache design determines how much context fits in memory. At the agent harness layer, harnesses such as Claude Code pick the context sent to the LLM. Prior work on world models has mostly focused only on the model layer. In this paper, we make three contributions. First, we build a benchmark to measure world model context retrieval. It generates new worlds for every episode, preventing a model from recalling rooms using training data. It also separates context retrieval from guesses based on nearby frames. Second, we build an agent harness that performs context compaction for world models, which to our knowledge no prior work has studied. Third, we use our benchmark to study four world model designs. Context compaction lets our world models retrieve rooms up to 4 times beyond their context window without retraining. However, world models with sliding-window attention lose most of their context retrieval when dreaming. Therefore, a dreaming world agent benefits more from extending the reach of its world model using context compaction than switching to sliding window attention at the model layer.

★ Beyond LLM Wrappers: Context Engineering and the Three-Engine Architecture for Reliable Financial AI

Pradeep Kumar GS, Ravi Tanniru — Fri, 30 Oct 2026 00:00:00 +0000

97% · Hallucination 97%, Agents 41%

Financial AI systems must support decision-bearing questions involving payments, collections, cash flow, and compliance while ensuring numerical accuracy, traceability, and auditability. Large language models (LLMs) enable natural language interaction but cannot independently guarantee reproducible calculations or evidence-grounded recommendations. We present a production-grounded architecture integrating three components: a financial knowledge graph for enterprise relationships, temporal context, and provenance; a deterministic optimization engine for financial calculations and decisions; and a multi-agent LLM layer for intent interpretation, task orchestration, explanation, and claim validation. A coordination layer, termed context engineering, selects, ranks, composes, and validates query-specific context across ERP systems, tools, and tenants. Based on observations from 15 deployments over approximately 11 months, single-agent pipelines exhibited hallucination rates of about 7-9% (5-10% across query types), including invented invoices and unsupported financial claims. Graph-backed multi-agent validation reduced observed rates to below 1%, although these findings are observational rather than causal. ERP data quality and tenant-specific schemas remained major implementation challenges. Reusable mappings, graph templates, and validation policies coincided with a reduction in deployment time from roughly six to two weeks. We argue that financial context engineering extends beyond text-centric retrieval-augmented generation by requiring structured relationships, temporal validity, exact numerical evidence, access controls, and validation against source records and solver outputs. We outline research challenges in evaluation, reproducibility, privacy, context quality, and cross-tenant generalization.

★ What Does "Now" Mean to Your Computer-Use Agent? Stale Screenshots, Page Loads, and Which Check to Run Before a Click

Vasundra Srinivasan — Fri, 30 Oct 2026 00:00:00 +0000

97% · Agents 97%, Evaluation 92%, Efficiency 38%

Computer-use agents plan on a screenshot for several seconds and then click at a position on the screen; if the page has changed in the meantime, the click lands on an unintended element. Prior work treats the planning interval as the source of this risk. Measuring planned clicks on 31 public web pages, we find that the risk lies in page loading: 0.4% of clicks planned on fully loaded pages became stale within 30 seconds, against 17% of clicks planned in the first second after a page opened. On these pages, waiting about one second for network activity to stop removes most of this risk, yet the wait rules of eight of ten widely used open-source agents leave more than 1% of clicks stale. Commercial pages keep loading for longer: in the HTTP Archive's public crawl, 15% of 65 desktop booking and buying home pages went 500 ms without network requests within 3 seconds of their HTML being parsed, against 63% of 3,028 .gov pages, and 52% shifted their layout within 6.5 seconds of a network-idle wait capped at 3 seconds, against 8% (61% against 26% among pages recorded in full). Where a check before the click is needed, as before irreversible actions and on pages that are still loading, a pixel comparison verifies the target's position as accurately as a frontier vision-language model at one eighth of the latency, and a 4B model matches the frontier model in verifying whether the previous action succeeded. We release a benchmark generator, a break-even formula for when a check pays, and three rules for agent developers: https://github.com/multimodal-ai-engineering/preflight-bench

★ Finding Is Not Returning, Returning Is Not Reconstructing: A Multi-Dimensional Evaluation of Agent Social Re-entry

Sheng Wan, Cong Ma, Sichen Dong, Qian'Ang Mao, Chenhan Zhang, Daning Hu — Fri, 30 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

As Large Language Model (LLM) agents increasingly participate in persistent social environments, understanding their long-term social interaction capabilities becomes increasingly important. While existing research has advanced long-term memory and conversational context retrieval, less attention has been paid to whether agents can autonomously re-engage with previously inactive conversations. In particular, remembering a past interaction does not necessarily mean that an agent can identify when to revisit it, decide to return, or reconstruct the context needed for a coherent continuation. To investigate this gap, we introduce agent social re-entry as a multi- dimensional evaluation problem, distinguishing three complementary capabilities: Finding, Returning, and Reconstructing. We develop a shared-snapshot evaluation framework that measures these dimensions independently, allowing us to examine how conversational resurfacing and persistent memory relate to different aspects of social re-entry. Evaluations across six LLM configurations show that resurfacing substantially improves target conversation identification, while persistent memory is associated with higher rates of autonomous target-directed replying. However, neither improvement is accompanied by consistent gains in contextual reconstruction. These results highlight the distinction between accessing previous conversations, choosing to re-engage, and maintaining conversational continuity. Our framework provides a way to evaluate these capabilities separately rather than relying on memory retrieval or conversational response quality alone.

★ H$^2$R: Hierarchical Hint-Based Reward for Mathematical Reasoning via Reinforcement Learning

He Tu, Siye Li, Hongyi Wen — Fri, 30 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, New model 92%

Reinforcement Learning with Verifiable Rewards has shown great promise in enhancing mathematical reasoning in large language models. However, existing approaches provide limited intermediate supervision for complex, multi-step reasoning tasks. To bridge this gap, we propose \textbf{Hierarchical Hint-based Reward}, a framework that leverages the intrinsic sub-question structure of mathematical problems as structured intermediate conditioning signals during RLVR training. Unlike chain-of-thought prompting, these hierarchical hints serve as logically grounded intermediate milestones that effectively reduce the reasoning horizon toward the final answer. To validate this framework, we introduce \textbf{HiMath}, a benchmark of hierarchically structured mathematical problems. Across diverse LLMs, hierarchical hints consistently improve reasoning performance even without task-specific training. When integrated into RL training via H$^2$R, they boost Qwen2.5-Math-7B accuracy on HiMath from 30.4\% to 39.1\% using hierarchical hints, and up to 45.6\% with Oracle Hierarchical Hints. This pronounced improvement demonstrates that incorporating reliable intermediate checkpoints into reward design is essential for sustaining complex mathematical reasoning.

★ EnvOS: Auditing Computer-Use Agents When the World Moves On : Phenomenon-Based Task Construction, Certification, and Evidence-Based Evaluation

Ankush Kadu, Athira Gopal, Niharika Singh, Arpit Jain, Aswanth Krishnan — Fri, 30 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 87%

Computer-use agents built on large language models (LLMs) work in software that can change while they act: an appointment moves, a policy is revised, a record is updated. Testing agents under such changes is harder than it looks. When an agent fails, the cause may be the agent, but it may also be a change that did not fire as intended, an update the agent never received, or a grader that scored the episode wrongly. We introduce EnvOS, a framework for building and certifying environments that change mid-task. Before any agent runs, EnvOS checks that each change fires correctly, that the task remains solvable, and that the grader separates correct from incorrect behavior; after each run, separate code re-checks the verdict from archived evidence. We apply it to stale information, where a value the agent has already seen changes before the agent acts on it. In clinic booking, Claude Opus 5 succeeds on 0/16 tasks when nothing extra is shown, 2/16 when reminded to re-check the source, and 15/16 when the updated value is shown; the gain over the reminder holds after correcting for all tests (p = 0.00293). In insurance-claim approval, Opus 5 re-checks on its own and succeeds on 10/10, whereas Claude Sonnet 5 rises from 1/10 to 9/10 when the current policy is shown again, a gain significant in its pre-registered test but not after correcting for all tests. Whether an agent caught a change thus varied with the model and the interface. We also stress-test the grading itself: on corrupted episodes designed blind from the rules, the EnvOS pipeline accepted none of 26 as clean, and on an externally generated set it caught 25 of 39, against 4 for simple log checks. Because every EnvOS environment is generated from a seed and graded by an audited automatic verifier, the same environments are natural candidates for training agents with reinforcement learning from verifiable rewards.

★ Beyond Selected Breakthroughs: Evaluating Language Models on Randomly Sampled Mathematical Conjectures

Helen Jenne, Davis Brown, Jesse He, Henry Kvinge — Fri, 30 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Hallucination 40%

The last year has seen a shift in AI-assisted mathematics, with frontier language models producing proofs of and counterexamples to long-standing open conjectures. However, studying only the current set of major breakthroughs provides an incomplete picture of LLM capabilities and the computational resources needed to achieve reported results. To address this, we introduce a new dataset, OpenConjecture, which collects conjectures from mathematics papers published on the arXiv since February 2026. We use these conjectures as a source of open mathematical problems, and apply frontier GPT models to a random sample of 200 of these conjectures over three passes using a reproducible protocol without problem-specific mathematical guidance. Rather than attempt expert verification of these candidate solutions, we analyze the types of problems for which LLMs produce solutions and the characteristics of these solutions. The models we test produce candidate solutions for roughly 60% of the conjectures, with counterexamples about 1.3 times more common than proofs. We estimate solution novelty, assess how solutions cite and relate to existing literature, and analyze whether the proposed resolution relies on either an error in or underspecification of the conjecture. We find that models sometimes propose resolutions that are outside the authors' intent (e.g., proposing the trivial group as a counterexample) and that a number of proofs follow closely from existing results, but that some resolutions exhibit at least moderate novelty. Our findings underscore the need for evaluations that go beyond correctness to assess fidelity to the intended statement, attribution, and the nature of the mathematical contribution.

★ Ramanujan Challenge Results: AI-Assisted Evaluation Pipeline for Multi-Format Proofs

Fri, 30 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Hallucination 72%

As AI's mathematical capabilities rapidly improve, meaningful benchmarks are shifting from university-level problems and math olympiads to research-level mathematics. The Ramanujan Challenge is a benchmark comprised of research-level problems in number theory, including unpublished results and open conjectures. It presented concrete formulas for mathematical constants whose correctness can be checked numerically, while their proofs may require non-obvious mathematics. Here, we analyze the Challenge submissions, measuring the relative success of different AI models, Computer Algebra Systems (CAS), and formal tools. We find correct submissions that are conceptually distinct from the official solutions, and a proof of one of the published open conjectures. Evaluating the submissions exposed a second challenge: judging complex proofs reliably at scale. We introduce an AI-assisted evaluation pipeline for natural language proofs, CAS code, and Lean formalizations that assesses semantic and logical correctness. Compared with the conventional one-shot LLM judgment, our pipeline reduces false-positive verdicts on natural language proofs from 23% to 6% and improves judgment accuracy on supporting code from 46% to 58%. This approach provides a methodology for evaluating research-level mathematics at scale, with potential applications across mathematics and AI for science.

★ WebAnything: From the Wild Web to Scalable Agent Environments

Tianci Xie, Zikun Cai, Chaoyun Zhang, Pinxue Guo, Bingxuan Li, Rui Zheng — Fri, 30 Oct 2026 00:00:00 +0000

96% · Agents 96%, Evaluation 76%

Web agents require diverse, executable, and resettable website environments for reproducible training and evaluation. Human-assisted mirrors demand per-site engineering, while generative construction can diverge from specified real websites. Capture-and-replay preserves observations, with interaction coverage constrained by captured content and replay capabilities. We present WebAnything, a fully automated framework that reconstructs source-grounded, reusable Web environments from real URLs without per-site manual implementation or repair after initial setup and authorization. An Explorer-Builder collects browser-observable structures, resources, and interaction outcomes; an independent verifier checks frozen candidates and guides local repair or targeted source recapture. We construct 244 website artifacts across 14 application domains and 53 categories, supporting 12,984 executed and accepted tasks. Compared with human-assisted WebHarbor, WebAnything achieves higher aggregate scores on three fidelity metrics, with 12 wins, two ties, and no losses in a blind audit. Fine-tuning Qwen3.5-9B on 5,856 interaction trajectories improves REAL success by 9.9 percentage points, Odysseys-Easy Rubric Avg and Task Avg by 6.25 and 6.32 percentage points, and WebArena-Verified Overall success by 13.33 percentage points under the fixed harness. Code: https://github.com/china-qijizhifeng/Web-Anything

★ Measuring Where Intent Gets Lost in Agentic Smart Contract Synthesis

Abhinav Goel, Agostino Capponi, Alfio Gliozzo, Chaitya Shah — Fri, 30 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Hallucination 40%

AI agents can now write software from a plain-English description, but code that runs is not the same as code that does what was asked. This gap matters most for smart contracts, the programs behind Web3 applications that move money on a blockchain and cannot be fixed once deployed. We ask how to check whether an AI-written contract behaves as it should, and where the requested behavior gets lost. We design an evaluation built on execution rather than opinion. We replay the same transactions on an AI-written contract and on the original it should reproduce and measure how often they agree. We test every grading tool on planted mistakes, compare against an answer key written without the description, and audit our own scorer until known wrong answers cannot pass. Common methods, such as AI judges and standard analysis tools, miss obvious mistakes, and scoring choices matter too. Checking only the functions every system includes hid about half the difference between pipelines. We then propose an elicitation-first architecture. Before writing code, the agent asks the client a short checklist of questions, checks the contract against an independently written executable answer key, and returns disagreements to the client instead of silently rewriting code. Asking first matters far more than reviewing or patching afterwards, across three AI models, while automated security fixes often break requested behavior. A few well-chosen questions recover much of what a short description omits, and their value can be estimated without the original contract. To our knowledge, this is the first study in smart contract synthesis to gather missing requirements before generation and check the result against an independently written executable model. The lessons extend to any setting where AI agents turn requests into programs. We release our tools, 600 tasks and all results.

★ Bi-Phase Training: Learning Efficiently From Scratch

Abel Gurung, Joseph Campbell — Fri, 30 Oct 2026 00:00:00 +0000

96% · New model 96%, Efficiency 30%

Pre-training foundational models is computationally expensive as models scale. In this work, we introduce Bi-Phase Training (BPT), a parameter-efficient pre-training method that preserves the high-rank learning dynamics observed in full-parameter optimization with substantially reduced trainable parameters. BPT parameterizes each weight update as a composition of constrained high-rank left/right transformations via trainable diagonal matrices and low-rank adapters whose updates are periodically merged into the frozen base weights, resulting in a compounding expansion of accessible directions during training. Across language-model pre-training from scratch (Qwen2.5 models from 100M to 1.5B parameters and OLMo-2-1B) on FineWeb-Edu (10B and 100B tokens), BPT achieves validation perplexity close to full-parameter baselines while using far fewer trainable parameters; e.g., at the 1.5B scale, BPT uses 66% fewer trainable parameters with only a small validation-perplexity gap. Additionally, a comprehensive evaluation across diverse downstream tasks shows that BPT preserves much of the full-parameterized model's downstream performance.

★ Nemotron 3.5 Super and Lightning: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Models for Agentic AI

Fri, 30 Oct 2026 00:00:00 +0000

96% · New model 96%, Efficiency 96%, Agents 80%

We introduce Nemotron 3.5 Super and Nemotron 3.5 Lightning, open and efficient Mixture-of-Experts Hybrid Mamba-Transformer models designed for reasoning, long-context understanding, and agentic workloads. Super is a 120B-parameter, 12B-active multimodal model, while Lightning is a smaller 30B-parameter, 3B-active text model; both designed for efficient deployment. Both models combine large-scale distillation, long-context training, and native speculative decoding, while Super further incorporates Indic continued pretraining and agentic software-engineering mid-training. We develop a comprehensive post-training pipeline spanning supervised fine-tuning and reinforcement learning (RLVR and RLHF), complemented in Super by specialized teacher training and multi- teacher on-policy distillation (MOPD). This transforms Super from a text-only base model into a multimodal agent capable of reasoning over text, images, video, and interactive tool environments. We further explore test-time scaling through parallel agent execution and self-verification. Across training and deployment, we jointly optimize capability and efficiency. Indic continued pretraining improves Indian-language benchmarks by up to 9.82 percentage points, while agentic mid-training strengthens software-engineering capabilities and robustness across agent harnesses. The post-trained Super model achieves 51.78% on SWE-bench Pro, surpassing Nemotron 3 Ultra despite using approximately 4.6x fewer active parameters. Post-training MTP boosting improves speculative- decoding acceptance length by up to 21.8% under probabilistic draft sampling, while NVFP4 quantization achieves 99.62% median accuracy recovery relative to BF16. Inference optimizations enable up to 4,856 output tokens/s/GPU on GB200 while sustaining at least 50 tokens/s/user.

★ Scaling Lean Theorem Proving for Real-World Code Verification

Fri, 30 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

We present EasyProver, a Lean proving system built from recursive agents with isolated contexts, asynchronous search, parallel tactic probing, and an elastic elaboration service. Using an off-the-shelf frontier model with no additional training, it proves 669/672 PutnamBench problems and all six IMO 2026 formalizations and resolves 597/600 NTP4VC verification conditions and all 189 VERINA tasks by proof or refutation. Because these benchmarks are close to saturation, we also introduce Aegis, which converts competitive programming problems into much harder verification tasks over real implementations. The pipeline writes safe Rust solutions, translates them into Lean with Aeneas, generates specifications, and labels each frozen statement with a kernel-checked proof or refutation. It produces 2,045 differentially validated Rust-Lean program pairs and 11,533 candidate specifications. Unlike hand-written benchmarks, these tasks carry implementation-level obligations from fixed-width arithmetic, explicit failure, generated loops, casts, and functionalized mutation. On a stratified sample of 2,000 statements, EasyProver proves 1,054 (52.7%) within a 30-minute budget per statement. The same tasks can also be used to benchmark other provers and as rewards for reinforcement learning.

★ Component Translation and Judge-Labelled Refusal: A Multilingual Safety Evaluation

Yongxi Zhou, Junwei Yao, Wenbo Ye, Jiahui Ge — Fri, 30 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

Whole-prompt translation evaluates a request in different languages but does not compare which component carries a language. We introduce a paired reversal: translate the complete request component or the complete operation component into the same target language, holding behaviour, English frame and translated-span count fixed. Component identity and non-English text quantity change together. Assigning English, Chinese and Spanish independently to the two components yields nine configurations on 190 of 300 sampled harmful behaviours retained after consensus annotation and automatic gating. Refusal scored by language-model judges is the primary outcome; unsafe compliance is reported separately. On Gemini 2.5 Flash and DeepSeek, three refusal swaps are about 5.3 percentage points and the fourth is -0.5; none survives the pre-specified six-test Holm correction. In an exploratory extension, Mistral Small 4 has a 13.7-point Spanish swap (95% interval [8.4, 18.9]), which survives the twelve-test Holm sensitivity correction under the primary judge, and 12.1 points under a second judge. GPT-5.4-mini estimates are within 2 points of zero. Component assignment supplies an additional evaluation dimension. Evaluations should record component languages and translated text quantity, comparing matched behaviours at equal translated-span count while interpreting refusal and unsafe compliance separately.

★ GAAC: Lossless KV Compression for Bandwidth-Efficient LLM Inference on Personal Devices

Yipin Guo, Arun M George, Siddharth Joshi — Fri, 30 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, Evaluation 32%

Weight-only, lossy quantization substantially reduces the storage and memory requirements for serving large language models (LLMs), but the growing key-value (KV) cache remains a bottleneck for long-context inference on personal devices. We present Group-Adaptive Additive Codec (GAAC), a lossless KV compression and serving system that exploits exponent redundancy to reduce storage and accelerate attention without introducing KV approximation. GAAC encodes BF16 exponents using calibrated per-channel base-plus-offset codes over contiguous ranges. GAAC adapts code widths across small groups while preserving KV values bit-for-bit. A fused attention kernel reconstructs KV tiles on the fly, avoiding full-cache decompression. Block-wise compression and fixed-capacity slots support incremental cache updates and paged serving. Across five tested models, GAAC achieves 1.405-1.407$\times$ lossless BF16 KV compression. We also study compression performance on low-precision tensors FP8 E4M3 caches ( 1.09-1.14$\times$) and E5M2 caches (1.25-1.30$\times$). On NVIDIA DGX Spark, GAAC accelerates decode attention at a 64K-token context by 1.19-1.27$\times$ and end-to-end W4A16 decode steps by 1.04-1.15$\times$. In serving, it reduces resident KV memory by approximately 26% and increases aggregate throughput by 13-29% under comparable KV-memory budgets. On Apple M4, GAAC achieves 1.02-1.25$\times$ attention speedups, demonstrating portability beyond NVIDIA GPUs.

★ SatOps-Decision: Low-Latency Discriminative Decisions for Onboard Satellite Communication Operations

Lifan Mei, Yuxuan Li, Jiaxian Tu, Shenghe Xu — Fri, 30 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Efficiency 90%, New model 48%

We propose SatOps-Decision, a method for low-latency decisions in satellite communication operations. To the best of our knowledge, it is the first application of Jev-style general-purpose discriminative decision models to satellite operations, and the accompanying benchmark is the first dedicated to this approach. The learned scorer, a locally adapted Laya encoder with its native decision head, selects among supplied operational answers in a single forward pass. The benchmark combines authored operational contracts, controlled state changes and independently checked labels in an auditable construction. Evaluation prioritizes complete-request latency and raw decision accuracy; all main accuracy comparisons use the same 1,024 held-out questions, separate from training and development. SatOps-Decision's local scorer has a median latency of 90.3 ms per decision, versus 493.9 ms for Qwen3.5-0.8B generation and 388.4 ms for OneJev1.6-0.8B: $5.47\times$ and $4.30\times$ speedups based on median request times in the measured FP32/MPS implementations. Its mean raw accuracy is 85.3%, versus 42.5% for frozen Laya, 40.7% for Qwen generation and 58.6% for OneJev. As cloud accuracy references, original Jev 1.14/1.13 reach 87.4/81.9%.

★ Aligning Autoresearch: Co-Improvement between Agents and Humans

Fri, 30 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, New model 84%, Agents 52%

Autoresearch agents can now attempt long-horizon AI research end-to-end, i.e., self-improvement. However, best results are still obtained by research agents interacting with human researchers, and improving AI's ability to perform this co-research will only reinforce this effect---a process termed co-improvement (Weston and Foerster, 2025). Further, aligning with human expectations is likely a safer path for research agents, especially given recent agent reward hacking phenomena (OpenAI, 2026b). In this work we address these issues in four ways: (i) we empirically study the role of humans in directing autoresearch agents, focusing on the domain of benchmark creation, and find humans still play a pivotal role particularly in the ideation phase; (ii) we also develop a benchmark, HackBench, that shows current autoresearch agents are not aligned and are unable to achieve hack localization sufficiently; (iii) we present a method to train agents for improved human interaction by using human simulators as a proxy during reinforcement learning (RL); and finally (iv) we develop a method to improve model text generation, and hence agent communication, by RL with rubrics aligned with human experts. Together, our training and evaluation recipes provide a path toward more capable and aligned (co-)research agents.

★ Efficient Agentic Audio-Visual Question Answering through Toolset Pruning and Modality Consolidation

Yi Wang, Tianzi Wang, Peter Bell — Fri, 30 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, Agents 87%

Fine-grained audio-visual question answering (AVQA) requires models to resolve complex cross-modal correspondences and temporally grounding from sparsely distributed evidence. Omni-modal agents, which autonomously orchestrate perception tools to actively collect evidence, achieve strong performance, but it remains unclear whether active perception from both audio and visual modalities depends on broad specialized toolsets, especially audio or visual modality-specific tools. Such systems may also introduce long execution trajectories and substantial inference cost. We investigate whether comparable performance can be retained with a compact workflow exposing only essential perceptual capabilities. Starting from the specialized-tool OmniAgent, we conduct staged controlled experiments that prune the toolset and progressively consolidate modality-specific tools into omni-modal tools, motivating a two-tool agent comprising only Omni-modal captioning and Omni-modal QA. Across three widely used AVQA benchmarks, our compact agent achieves comparable or better accuracy than the specialized-tool agent while reducing inference cost by around 70%.

★ What Breaks When Facts Become Stories? The Distortions Source-Agreement Metrics Miss When Models Retell Real Events

Kwangsuk Park, JiWoong Yang, Vaniushkina Kseniia — Fri, 30 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Hallucination 71%

Language models are now widely used to turn recorded facts into media for public consumption, yet evaluation of such output still assumes agreement with the source as its measure of success. In reconstruction, where departing from the source can itself be the goal, this criterion measures the wrong construct: applied as is, it counts intended departures as errors; withheld, it lets real distortion through. We had three model families rewrite 2,000 news articles as summaries and as stories, measured what changed on every axis of the five Ws and how, and asked of each detected distortion whether the assigned perspective warranted it, which set a third of them aside. Time and manner were not lost but replaced, while place and causation lost the expressions that carry them. Accountability distortion behaved otherwise. Barely present in summaries, it rose up to sevenfold in stories, and of those whose direction could be determined, two in three lightened the perpetrator's responsibility. Such distortions survive claim-level fact checking and leave the retelling grammatically cleaner than its source. Where our judges disagreed, three authors relabeling the same cases converged on whether a distortion had occurred but not on whether it was warranted, a boundary contested rather than merely hard to detect. To support work on detecting and reducing them, we release ReTold-2k, a reconstructable corpus with the scoring protocol and per-judge baselines kept unaggregated.(https://github.com/Retold2K/retold2k).

★ From Intent to Improvement: Agentic Dataset Search for Better Model Training

Qiaosheng Chen, Kaijia Huang, Fanqi Zhou, Gong Cheng — Fri, 30 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Agents 87%, Hallucination 56%

Finding training data in million-scale machine learning repositories is a Web search problem where text relevance is not enough. A query couples a semantic training goal with constraints such as task, license, and training stage, and the metadata that would verify them are often missing. Text retrieval ranks dataset cards by similarity but cannot enforce hard constraints or follow the relations that reach a sparsely described dataset. Free-form tool-calling agents are more flexible, but the same model that picks a dataset also declares it a match, and nothing checks that the query's constraints are actually met. Neither approach establishes that the data it returns will improve model training. To address it, we propose AgenticDS, a contract-governed agentic framework in which every language model decision instantiates a typed object before deterministic code consumes it. The framework compiles a query into a typed search plan, executes it through admissible deterministic operators over hybrid retrieval and bounded graph expansion, and verifies the provenance-grounded evidence before accepting a ranking or emitting a typed diagnosis that permits only local repairs. On HuDS-Bench, a new benchmark of 200 queries with 24,908 graded judgments pooled over 15 systems, AgenticDS improves NDCG@10 by 25.7\% over the strongest dense--sparse hybrid and by 18.9\% over a commercial LLM agent. A controlled search-to-train study further shows that the datasets it retrieves give the best downstream accuracy in mathematics, code, and medicine. Code and data are publicly available at https://github.com/nju-websoft/AgenticDS.

★ DWISE: Denoising Workload-Informed Scheduling Engine

Lie Li, Wen Li, Junxiao Shen, Jindong Gu, Guosheng Hu — Fri, 30 Oct 2026 00:00:00 +0000

96% · Efficiency 96%

Unlike autoregressive large language models (AR-LLMs), diffusion LLMs (dLLMs) generate text by iteratively denoising many tokens in parallel, offering a promising path to efficient inference. Existing serving frameworks, however, schedule dLLM requests with policies designed for AR models, namely first-come-first-served (FCFS) admission and prefill priority. We find that both policies are ill-suited to dLLMs, wasting much of their speed. FCFS ignores that a dLLM request's decoding cost is highly heterogeneous and unknown at arrival, causing severe head-of-line blocking. Prefill priority, widely used in AR-LLM serving to keep time-to-first-token low, sharply cuts dLLM efficiency, as rounds that run only prompts leave most of the batch idle. We present DWISE, a Denoising Workload-Informed Scheduling Engine for efficient dLLM serving. DWISE schedules requests by their remaining work, measured in a block-level unit shared by prefill and denoising, so both phases share a round. To predict the unknown denoising workload, DWISE combines a lightweight prompt-based predictor at admission with a single refinement from the first denoising round's hidden states, which capture model-specific information absent from the prompt and require no additional model execution. Implemented in SGLang and evaluated on five models from three dLLM families under bursty workloads, DWISE reduces mean end-to-end latency over SGLang by 5.4$\times$ on average and up to 14.9$\times$ while preserving task accuracy. The code will be released publicly.

★ Multi-Draft On-Policy Distillation for Diffusion Drafters

Petr Molodyk, Yuchen Zhu, Jaemoo Choi, Molei Tao, Yongxin Chen — Fri, 30 Oct 2026 00:00:00 +0000

95% · New model 95%, Efficiency 92%

Speculative decoding is used to reduce the inference cost of language models. One-step diffusion drafters propose a block of multiple tokens in parallel, reducing the sequential cost of speculative decoding. On-policy distillation trains these drafters using verifier feedback on their own proposals. However, parallel drafting makes this feedback difficult to use: the drafter predicts each token without observing the preceding tokens sampled within the same block, while the verifier conditions its prediction on those tokens. Different sampled prefixes therefore give different verifier targets for the same drafter prediction. We show that distilling against these targets separately can suppress tokens that are plausible after some prefixes but unlikely after others, harming draft acceptance. We explain the effects of this mismatch and introduce Multi-Draft On-Policy Distillation (MDraft), a new parallel drafter distillation method which samples several draft blocks from the same context and aggregates the verifier's full-vocabulary conditionals across their prefixes. We use all the drafts to collect information from the verifier, allowing support from alternative prefixes to influence the training target. The resulting algorithm improves the acceptance length of a parallel drafter beyond the level of pre-training, and alleviates the effects of the distributional mismatch. MDraft changes only training, retaining the drafter's architecture and inference procedure, allowing the resulting drafter to be used as-is in place of the pretrained checkpoint. Experiments on a suite of mathematical, coding, and conversational tasks show longer accepted drafts than prior distillation methods.

★ Think Twice: The Metis-1.6 Technical Report

Giulianno Vollmer — Fri, 30 Oct 2026 00:00:00 +0000

95% · New model 95%, Efficiency 67%, Evaluation 34%

Metis is a bet that better data, better training methods and new architectures can buy what a bigger parameter count cannot, at a fraction of the cost. Metis-1.6 is its latest step: two language models built from scratch by a single researcher, Praxis (2.94B parameters, 0.46B active per pass) and Logos (10.23B parameters, 1.18B active per pass). Both run a hybrid Mamba-2 and attention mixture-of-experts stack twice for every token, so each token gets a second look. Each was trained on 1.02T tokens in total: 804.76B of pretraining, 18B of context extension to 131,072 tokens, and 200B of supervised fine-tuning. Metis-1.6 was never meant to challenge frontier models, and across 40 post-trained benchmarks it mostly does not: Logos averages 52.93, seventh of 17 models in its comparison set and within a point of Qwen3-4B. But it gets there on roughly a twelfth of the median training data of the comparators that disclose it, and with fewer active parameters per token than any model that beats it, which puts it on the Pareto frontier for both. In mathematics it ranks second of 30 models, with 67.50% on AIME 2025 and 63.64% on HMMT February 2026. It is weak at code, instruction following and long context. The rest of the report is the part I found most valuable: what went wrong. The optimizer I configured never ran, and attention trained with AdamW for all of pretraining. A 1.77B-parameter n-gram memory learned to switch itself off. Five fine-tuning choices cost capability. And a length reward under group-mean normalization punished correct solutions for being long once a group's pass rate passed 2/3, a sign inversion I prove and fix with a one-line clamp. I close with the 10B-token ablations behind MoRE (Mixture of Recursive Experts), the looped architecture of Metis-1.7.

★ Understanding Large Language Models as Numerical Optimizers

Lingyun Guo, Ali Bereyhi — Fri, 30 Oct 2026 00:00:00 +0000

95% · Evaluation 95%, Hallucination 39%

Large language models (LLMs) have emerged as potential tools for numerical optimization. Despite promising results, their intrinsic optimization capabilities and limitations remain poorly understood, particularly in high dimensions. This paper investigates the ability of LLMs to exploit mathematical problem structures and generate feasible, near-optimal solutions. We develop an evaluation framework that distinguishes solution quality from execution failures, including numerical inaccuracies, constraint violations, and output-format errors. Using representative problems from convex optimization, fractional programming, and constrained projection, we assess LLM performance against deterministic reference solvers across different dimensions. Our results reveal that LLMs can capture useful optimization structures and produce near-optimal solutions, even when their outputs fail to satisfy numerical or formatting requirements. However, reliability deteriorates with increasing problem dimension and depends strongly on the underlying optimization structure. To further investigate the limitations, we deploy an iterative refinement mechanism that provides targeted feedback on feasibility and solution quality. The results demonstrate that iterative refinement substantially improves reliability and extends the range of problem dimensions over which LLMs produce valid solutions. Nevertheless, persistent numerical and structural errors limit their applicability as standalone optimizers. These findings provide insights into the capacity and limits of LLM-based optimization and motivate hybrid approaches that combine LLM reasoning with deterministic numerical verification.

★ SkillMem: Compact External Skill Memory for Latent Multi-Agent Reasoning

Yuxuan Zhu, Mohammad Mohammadi Amiri, Pin-Yu Chen, Keerthiram Murugesan — Fri, 30 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, New model 53%, Evaluation 37%

Latent multi-agent systems reduce inference cost by keeping intermediate reasoning and inter-agent communication in continuous representations.\ Reusable skills, however, are commonly supplied as text or stored as layer-wise Key-Value states or skill-specific model parameters, introducing repeated processing or substantial per-skill storage.\ We introduce \textbf{SkillMem}, which represents each reusable skill as a small external latent memory and makes it directly available during latent multi-agent reasoning.\ A role-conditioned cross-attention reader allows the same memory to guide different agents according to their roles and current hidden states.\ We train the skill memories and shared reader through behavioral distillation, matching the change in token predictions induced by a textual skill while keeping the backbone frozen.\ For our experiments, we obtain textual skills from paired successful and failed multi-agent trajectories, although existing skills can also be compiled directly.\ Across six benchmarks and Qwen3 models ranging from 4B to 14B parameters, SkillMem provides strong gains for smaller backbones while preserving efficient latent multi-agent inference across model scales.\ We use the same $8\times256$ memory configuration across all evaluated model scales, requiring 9.6\,KB of skill-specific storage per skill, including its retrieval embedding.\ For a library of 100 skills, SkillMem reduces total storage by more than $38\times$ compared with alternative skill representations and achieves up to a $4.23\times$ end-to-end inference speedup.\ These results show that compact procedural memories can be reused without sacrificing the efficiency of latent multi-agent reasoning.

★ SkillCT: A Multidimensional Benchmark and Execution-Free Baseline for Agent Skill Assessment

Mengyu Sun, Zunlong Zhou, Ziyuan Yang, Yi Zhang — Fri, 30 Oct 2026 00:00:00 +0000

95% · Evaluation 95%, Efficiency 83%, Agents 72%

Agent skills package instructions, scripts, and resources for specialized tasks and are increasingly distributed through public Web repositories. However, existing skill benchmarks primarily focus on downstream task execution, insufficiently assessing specification quality, behavioral robustness, semantic alignment, and responsibility boundaries. Their reliance on task-specific tools, environments, and repeated executions limits scalability and pre-deployment assessment. To address these challenges, we introduce SkillCT, a benchmark that broadens skill assessment from downstream utility to a multidimensional characterization of skill quality. SkillCT evaluates skills across nine complementary dimensions spanning three aspects: Static Assessment, which captures intrinsic properties of the skill specification; Dynamic Assessment, which measures whether the induced behavior remains stable across backbone models and request variations; and Human-Assisted Assessment, which examines behavior-chain consistency and task granularity. These dimensions provide a broader view of skill quality than execution success. Applying SkillCT to a public skill corpus establishes a reference benchmark for analyzing quality weaknesses and relationships across dimensions. Building on this benchmark, we propose an execution-free quality evaluation baseline that predicts the multidimensional quality profiles of unseen skills by retrieving and leveraging similar reference skills, without executing the skills under evaluation. This reduces token consumption and evaluation time, while preserving consistency with the full assessment. Furthermore, SkillCT-guided optimization improves downstream task success and execution efficiency, demonstrating its value for skill refinement. These findings highlight the effectiveness of SkillCT in supporting scalable skill assessment, optimization, and repository maintenance.

★ Multi-Token Prediction Done Right

Jayden Teoh — Fri, 30 Oct 2026 00:00:00 +0000

95% · New model 95%, Efficiency 91%

Multi-Token Prediction (MTP) is now a common ingredient in language-model pre-training, as it improves downstream performance and enables faster inference through speculative decoding. We identify a flaw in standard MTP training, in which a residual component of the MTP gradient interferes with both next-token and multi-token learning, increasingly so with scale. Specifically, the MTP cross-entropy gradient can be decomposed exactly into a KL self-distillation term that pulls each MTP head toward the model's own next-token prediction (NTP) distribution and improves speculative decoding, and a residual that has a low signal-to-noise ratio and buries the useful KL signal. We introduce the KL gradient-recovery trick: when two distributions share a linear output head, the gradient of their KL divergence can be recovered exactly by combining their cross-entropy gradients through straight-through gradient routing. Unlike minimizing the KL directly, this trick requires no vocabulary-sized logit matrices and is compatible with standard fused cross-entropy kernels. This enables Soft Multi-Token Prediction (Soft-MTP), a one-line code change to MTP that removes the noisy residual and backpropagates only the KL self-distillation signal by reusing the existing NTP and MTP cross-entropy computations. Soft-MTP supports both DeepSeek- and Meta-style MTP, adds no parameters, and incurs no measurable overhead. From 167M to 1.2B parameters, Soft-MTP improves over standard MTP with lower NTP perplexity, faster speculative decoding and higher zero-shot accuracy at every scale, lower MTP perplexity from 689M onward, and stronger math and code reasoning after fine-tuning at 1.2B. Soft-MTP also matches, if not exceeds, direct KL divergence minimization on most metrics. Soft-MTP is, in effect, a free lunch for MTP optimization. Our code can be found at https://github.com/JaydenTeoh/Soft-MTP.

★ SCOPE: Selective Critical-Step On-Policy Self-Distillation for Agentic Reinforcement Learning

Qirui Liu, Wei Cao, Yu Mi, Yichen Sun, Xuekun Zhang, Lei Liu — Fri, 30 Oct 2026 00:00:00 +0000

95% · Agents 95%, New model 93%, Efficiency 55%

Outcome-based reinforcement learning provides coarse supervision for long-horizon LLM agents: failed trajectories may contain correct steps, while successful trajectories can contain errors. Trajectory-level skills can also shift teacher preferences at unrelated steps, introducing off-target distillation signals. We introduce SCOPE, a framework for Selective Critical-Step On-Policy Self-Distillation. After a supervised warm-up, the policy analyzes its own trajectories, generates step-specific hindsight skills, and distills their guidance only at selected error steps. A bounded signed teacher-student coefficient stabilizes auxiliary updates while preserving positive reinforcement and negative correction. A success-relative pairwise length reward further encourages efficient task completion. Experiments on ALFWorld and WebShop demonstrate consistent improvements across multiple model backbones. With Qwen2.5-3B, SCOPE achieves 97.7% success on ALFWorld and 85.2% on WebShop, outperforming SEED by 5.9 and 6.3 percentage points, respectively. On unseen ALFWorld tasks, SCOPE reaches 89.6% success, improving over GRPO by 23.2 percentage points with 46.7% fewer interactions. Scaling to Qwen3-32B, SCOPE achieves 99.2% success on seen ALFWorld tasks and 94.0% on unseen tasks.

★ Does Chain-of-Thought Reliance Follow a Scaling Law?Larger Models Follow Their Steps More, Independent of Post-Training

Esther Xin — Fri, 30 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

Monitoring a model's chain of thought (CoT) is useful for safety only if the final answer depends on the written steps. Prior measurements use natural tasks where true reasoning is unknown. We measure it against an exact ground truth on a synthetic box-swap task where every intermediate state is known. We replace one step inside a model's own chain with a wrong but valid state and check whether the answer follows the edit. We evaluate 25 open-weight models--Qwen3 (0.6B-14B, base and post-trained), Gemma 3 (1B-12B), Llama 3 (1B-8B), and 8 OLMo 2 checkpoints--on 7,200 items each. We find five results: (1) Reliance rises with size and saturates. In Qwen3, answer reliance rises from 16-18% at 0.6B to 97-100% at 14B; a logistic in log N fits well (R^2 = 0.89), and Gemma 3 and Llama 3 follow this trend. (2) Capability is the main driver. Across all models, reliance and step accuracy correlate at r = 0.99; equally capable models are equally reliant, regardless of family. (3) Size still matters at matched correctness. On problems solved correctly before the edit, reliance increases with size among capable models (rho = 0.80, p = 0.002). (4) Post-training does not change reliance. Base and post-trained Qwen3 differ by at most 0.03 from 4B upward; a large gap at 1.7B matches a large capability gap. (5) No model up to 14B solves the task internally. Without written steps, every capable model is at chance once a problem has four or more swaps. OLMo 2 checkpoints never become capable, making the training-token axis a limitation. On tasks requiring step-by-step computation, scale makes the chain of thought more load-bearing, supporting CoT monitoring in this regime.

★ What Hallucinations Have We Measured? Semantic Illusion, Confabulation, and Expressed Uncertainty

Fri, 30 Oct 2026 00:00:00 +0000

95% · Hallucination 95%, Evaluation 83%

Hallucination remains one of the most critical challenges in deploying large language models (LLMs) reliably and safely. Yet the term is used inconsistently across the research community. We argue for a strict definition of hallucinations in LLMs. First, hallucinations should be confident incorrectness, kept separate from expressed uncertainty (UNK)1. UNK reveals that the model may be wrong, whereas a hallucination does not reveal it and is therefore more dangerous in practice. Conflating the two obscures their distinct behaviors and makes them hard to study. Second, hallucinations should be split by a simple re-elicitation test: does any semantically equivalent prompt recover the correct response? This test probes whether the relevant knowledge2 is accessible. If none does, the knowledge is likely absent or inaccessible, and the error type is a confabulation (CF) that likely requires external knowledge or retraining; if some does, the knowledge is accessible from the model but applied inconsistently, and the error type is a semantic illusion (SI) that simple prompt or decoding changes can mitigate3. Re-evaluating 14 recent hallucination methods under this taxonomy yields three findings. (1) Detection methods flag UNK more often than hallucinated responses (i.e., SI and CF). (2) Mitigation methods are far more effective for SI than for CF. (3) Latent signals are more informative than output-level signals for CF. An audit of recent hallucination papers shows that most conflate these categories. We recommend reporting CF, SI, and UNK separately, and introduce a practical labeling protocol that needs only a modest number of semantically equivalent prompts.

★ Rethinking LLM Jailbreak Evaluation: Continuous Risk Landscapes Beyond Binary Outcomes

Yukai Zhou, Wenjie Wang — Fri, 30 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

A binary success indicator is a natural representation for discriminative tasks such as image classification under adversarial perturbations, where each input is mapped to a single prediction. However, generative tasks are fundamentally different: an LLM jailbreak prompt induces a distribution over open-ended trajectories, while a sampled response reveals only one path. Yet LLM jailbreak evaluation typically retains the binary response indicator underlying attack success rate, obscuring the surrounding structure of possible generations. We introduce a continuous, context-conditioned risk landscape over first-token branches, estimated by fixing candidate openings and repeatedly sampling their continuations. This continuous risk-landscape perspective reveals two complementary properties: current decoder exposure and safe rerouting capacity, which have been obscured by binary-indicator-based ASR. The resulting representation also provides a common basis for comparing clean prompts, attacked variants, and attack strength against a known adaptive defense. Across multiple target models and jailbreak attacks, we uncover risky branches beneath apparently safe outputs, identify qualitatively different ways attacks reshape the model's opening distribution, and show that attacks strongest before a defense do not necessarily remain the strongest after the adaptive defense reacts. These findings recast jailbreaks as deformations of the surrounding continuation-risk landscape, establishing a more informative foundation for understanding jailbreak behavior and developing safer language models.

★ Weaken to Strengthen: Weakening-Induced On-Policy Self-Distillation for Large Reasoning Models

Yecheng Wu, Song Han, Han Cai — Fri, 30 Oct 2026 00:00:00 +0000

95% · New model 95%, Evaluation 55%

Self-distillation offers a way to improve reasoning models when a suitable external teacher is unavailable. However, common approaches rely on privileged information to construct a stronger self-teacher, limiting their use when only unlabeled prompts are available. In this paper, we introduce Weakening-Induced On-Policy Self-Distillation (WISD), which instead constructs a weaker counterpart of the original model and uses the contrast between the two as a training signal without requiring privileged information. We explore two ways to construct the weaker model: low-bit quantization and magnitude pruning. We evaluate WISD across diverse architectures and scales on mathematical and coding reasoning tasks. Across these benchmarks, quantization and pruning improve average accuracy over the original models by 4.0 and 5.5 percentage points, respectively, achieving state-of-the-art performance even without privileged information. Notably, pruning raises Qwen3.5-4B's accuracy from 57.7% to 70.5% on HMMT 2025 and from 39.1% to 56.7% on LiveCodeBench v6. Quantization raises Qwen3.6-35B-A3B's accuracy from 68.5% to 74.6% and from 61.8% to 69.5% on the same benchmarks, respectively. These results establish WISD as a novel and effective self-distillation framework for large reasoning models.

★ The Quine Test: How Much of a Model's Self Survives Its Own Description?

Bibin Babu, Ivan P. Yamshchikov, Aleksei Shpilman — Fri, 30 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

Frontier language models express coherent preferences, but behavioral evidence alone cannot say whether these belong to the model or to a character it portrays. Following Hofstadter's view of the self as a self-description, we ask whether a model's description of itself works as a portable specification of its preference profile, a quine test: can the text a model writes about itself regenerate its behavior elsewhere? Ten models from nine labs write self-descriptions; we place each one in the prompt of other models (receivers) and of fresh instances of the author, and measure how closely their choices on a 110-item preference battery match the author's. First, the tested native self-descriptions (native: the profile measured under a generic assistant prompt) show no average reconstruction gain across variations in length, framing, wording and prompt placement: on a 0 to 1 fidelity scale where a model's fidelity to a fresh instance of itself is 0.979 and to unrelated models about 0.89, the one-sided 95% upper bound on the pooled gain is below +0.01 in every tested system-prompt condition, and descriptions based on the author's observed choices do no better. Second, the channel itself is strong: explicitly listing the author's answers lets receivers reproduce its profile at fidelity 0.955 to 0.960 across three model grids, within 0.03 of same-model repeatability. Third, self-descriptions written under an explicit preference policy transfer the assigned preferences almost as well as the policy text itself. Fourth, reading its own self-description shifts even the author away from its baseline, modestly more than unrelated prose of the same length. For the preferences measured here, native self-descriptions influence behavior without reliably specifying how to reproduce it in another model. We release the pipeline, battery, raw call logs, and 198 model-written descriptions.

★ Learning to Reason from Reference Samples Alone: Reinforced Adversarial Learning for Modern LLM Post-Training

Fri, 30 Oct 2026 00:00:00 +0000

95% · New model 95%, Evaluation 67%

In many settings where LLM post-training is desirable, high-quality answers are recognizable but difficult to score with explicit rewards. Reinforcement learning with verifiable rewards (RLVR) is limited to tasks where an explicit correctness signal is available. Prior adversarial approaches to training reasoning LLMs either complement verifiable rewards with co-trained critics or require reference reasoning traces. We ask whether a co-trained LLM judge can turn reference answers alone into the sole reward for on-policy post-training of reasoning LLMs. We call this method Reinforced Adversarial Learning (RAL). We first validate RAL on three verifiable tasks (integer multiplication, code-based arithmetic-expression synthesis, and SMILES-to-formula translation), using the verifiers only for evaluation. With sufficiently capable judges and enough training, RAL matches direct RLVR. We then apply RAL to scientific abstract generation in two settings. In the first setting, RAL guides the generator to produce abstracts in the scientific domain of the reference abstracts, even though the prompt leaves that domain unspecified. In the second setting, we generate abstracts for U.S. National Institutes of Health grant proposals conditioned on the grant title. Human scientists and an ensemble of LLM judges find our generations harder to distinguish from real abstracts than those of frontier-LLM baselines. These results suggest a path toward post-training reasoning models in scientific and other high-value domains, where increasingly capable judges infer hard-to-specify reward functions from example answers.

★ Lev-3B: A Data Recipe for Compact Probabilistic Log Decision Models

Leo Betthauser, Yingqi Liu, Uday Shah, Jishang Wei, Sainyam Galhotra, Liang Gou — Fri, 30 Oct 2026 00:00:00 +0000

95% · New model 95%, Efficiency 67%, Evaluation 44%

Many log-analysis tasks involve choosing among known options, such as detecting anomalous sessions or identifying attack techniques, and benefit from calibrated probabilities. We present Lev-3B, an open 3B-parameter decision model built on Granite-4.1-3B-Base with a shared pointer readout, and study its training recipe. The recipe has two key parts. First, training follows a curriculum: the model learns bounded natural-language decisions, using no log data, before it sees any logs. Second, the log stage mixes the target tasks with general log data spanning 35 auxiliary tasks unrelated to either evaluation task. We evaluate one shared checkpoint on sources seen in training and on sources excluded from training, calibration and model selection. The two-stage recipe transfers substantially better than training on log data alone. Adding general log data improves performance on excluded sources. At a fraction of their serving cost, Lev-3B outperforms prompted frontier models on anomaly detection, including on an excluded logging system, but trails them on attack-technique classification. Its probabilities are well calibrated on familiar sources, but on the excluded logging system it confidently misses anomalies. We release the checkpoint, recipe, data manifests, and prediction-level artifacts.

★ XLaw-JP-27B: A Lightweight Open Model Specialized for Japanese Law

Andrew Shin — Fri, 30 Oct 2026 00:00:00 +0000

95% · New model 95%, Evaluation 70%, Efficiency 45%

We present XLaw-JP-27B, an open-weight model specialized for Japanese law that runs on a single 48\,GB GPU under the Apache~2.0 licence. It adapts Qwen3.8-27B with hybrid retrieval over all current statutes and 31k case summaries, a light QLoRA fine-tune on its own verified answers to past bar-exam questions (self-distillation), a contract-specific LoRA adapter, and four-sample voting. We evaluate it on eight Japanese legal tasks---the multiple-choice section of the bar examination in 2024, 2025 and 2026, statement-level bar questions (JBE-QA), statute questions (LawQA-JP), the patent attorney exam, and contract review and drafting built from government model contracts---against open models of 12--33B under identical prompts and retrieval. XLaw-JP-27B ranks first on six tasks, ties on one and is third on one (mean rank 1.3 against 2.8--3.1 for the baselines). It is the only system that passes the bar exam in all three years (129, 120 and 130 of 175, meeting every subject minimum; no test-year material was used in training), while open Japanese 32--33B models fail at least one year. A controlled study of each component shows that retrieval, voting and self-distillation help, and that thinking mode, retrieval-grounded synthetic tuning, self-verification, per-statement decomposition and more self-distillation do not. The largest remaining gaps to frontier models are in contract review and patent law. The model is available as a free web service at https://x-law.jp/ with the weights available at https://huggingface.co/X-LawJapan/xlaw-jp-27b under Apache~2.0.

★ Unlimited OCR Works 2

Shu Wei, Jingjing Wu, Qunyi Xie, Lingshu Zhang, Xiangcheng Du, Yannian Fu — Fri, 30 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, Evaluation 94%, New model 65%

Unlimited OCR ushers the era of long-horizon parsing, substantially mitigating the growth of working memory as document lengthens. However, as document length scales to hundreds of pages, the visual context grows substantially, sharply increasing the computational cost of each decoding step and limiting the scalability of ultra-long document parsing. Inspired by human book transcription via localized focus without re-scanning preceding pages, we introduce Unlimited OCR 2, a document parsing model that looks only at what it transcribes. Building on Unlimited OCR, we upgrade all the decoder attention modules to Sparse Reference Sliding Window Attention (SR-SWA), bounding active reference tokens to a fixed capacity and thus sustaining constant computational costs across arbitrarily long documents. By combining SR-SWA with Unlimited-Horizon Decoding Paradigm, Unlimited OCR 2 enables document parsing over virtually unlimited lengths without prohibitive memory overhead. At the hundred-page scale, Unlimited OCR 2 achieves 7.0$\times$ and 5.6$\times$ per-step speedups over Full Attention and R-SWA, respectively. Furthermore, it delivers state-of-the-art end-to-end performance on both MPDocBench and MarathonDocBench, our newly proposed benchmark targeting hundred-page document parsing. Ultimately, the capacity for uninterrupted, one-shot document parsing signifies a fundamental paradigm shift into the OCR 3.0 era.

★ How Specific Do I Need to Be? Effects of User Query Underspecification on Deep Research Agents

Yi Wen, Xiangjue Dong, Bingxuan Li, Hangxiao Zhu, Leo Zhicheng Liu, Yu Zhang — Fri, 30 Oct 2026 00:00:00 +0000

95% · Agents 95%, Evaluation 94%

Deep research agents (DRAs) conduct long-horizon reasoning, search, and literature synthesis to provide long-form answers to complex, domain-specific questions. These agents are typically trained and evaluated on fully specified user queries. However, real-world user queries can often be underspecified. We introduce UNDERSPEC-DR, a framework for systematically evaluating how underspecified queries affect agents' final answer quality and intermediate search behavior. Across three DRAs we evaluated, underspecification substantially reduces how well DRA's answers fulfill users' information needs, even though they remain grounded in cited evidence. Our analysis of intermediate search behavior suggests a possible explanation: DRAs conduct broader searches that draw less on the original user query but rely more on their learned prior during training, which can be misaligned with users' intent. Our findings reveal the vulnerability of current DRAs to underspecified user queries and inform future efforts to build DRAs that proactively recognize when additional user input is needed to reliably complete deep research tasks.

★ Can LLMs Ignore What They Know? Measuring Knowledge Leakage in Persona-Based User Simulation

Sangmin Lee, Sangwon Lee, Sungbeom Choi, Jeeseung Park, Kibeom Myoung — Fri, 30 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

When persona-conditioned responses are read as respondent-level simulations, each response should reflect the described persona's knowledge, and the respondent written to know nothing must act as if the model's knowledge were not there. We test whether it can, and find persistent favourable-item real-twin rating separation under the tested low-knowledge and explicit-ignorance prompts. We define the resulting problem (knowledge leakage: a response bias that arises when knowledge the real user could not have enters the simulated response) and present a methodology for measuring it quantitatively. The core instrument pairs real items with fabricated twins matched in surface form: screened fabricated variants designed to minimize prior item-specific knowledge. The paired difference therefore provides a diagnostic contrast for knowledge leakage even in rating tasks where no answer is correct. In a rating experiment across three knowledge-handling arms and five API endpoints from two model families, we find, first, that the personas we test do not fully suppress real-item existence and familiarity signals: a favourable-item real-twin rating contrast remains under an explicit in-context statement that the persona has never heard of the item. Second, the unfavourable-item contrast reverses sign across domain groups: pooling all unfavourable-content items across domains the contrast sits near zero (+0.077 [-0.057, +0.204]), yet the same estimator in the same conditions returns -0.359 where the rating scale asks about an item's documented hazard and +0.438 where it asks about a company or an organisation. A pooled analysis of unfavourable-content items against zero therefore does not detect the opposing domain-group differences. All estimates are conditional on the adequacy of twin construction and screening; human low-knowledge baselines remain necessary for population-level interpretation.

★ rTaste: Do Research Agents Have Taste? Evaluating Early Stage Research Judgment

Art Kanke — Fri, 30 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

Do research agents have taste, the judgment of which experiments are worth running before their outcomes are known? We give six frontier models seven recent NLP papers. Each model, working as a research agent alone or as a team, proposes minimum viable experiments (MVEs), forecasts whether each will pass a threshold it sets and ranks them. A separate executor then attempts every MVE, including those ranked last, and three model auditors judge the results. Of 415 MVEs, 194 are judged sound, valuable and new, but fewer than half of these (73) are also judged beyond the seed paper and non obvious. The agents' first choices match the auditors' most valued experiment more often than a random pick, but are not shown to improve on choosing the experiment written first. Their forecasts beat a base rate but are not measurably more accurate than those of a fresh instance of the same model given only the paper and registrations. Teams cost more without a clear overall improvement in proposal quality. The agents produce many worthwhile experiments, but the evidence that their rankings improve the choice of which to run remains limited. The evaluation harness and a dataset of all 415 MVEs, with their forecasts and audited outcomes, are available at https://github.com/ArtKanke/rTaste.

★ Every Token Needs Its Own Teacher: Multi-teacher Adaptive On-policy Self-distillation

Yifu Luo, Chuang Ma, Zeyu Chen, Enkelejda Kasneci, Qi Chen, Gjergji Kasneci — Fri, 30 Oct 2026 00:00:00 +0000

95% · New model 95%

On-policy Self-distillation (OPSD) has proven effective for post-training large language models by constructing a teacher conditioned on privileged references. However, existing approaches keep the teacher fixed either throughout training by consistently exposing it to the full reference, or within each response by adaptively varying partial reference exposure only across responses. In this paper, we argue that neither of them is sufficient. Instead, every token needs its own teacher, as the teacher preference can vary token by token within the same response. Through a teacher-intervention pilot study, we validate that the teacher preference differs across tokens, and reveal that a preferred teacher can be identified for each token through majority voting and KL-based selection. Motivated by these findings, we propose Multi-teacher Adaptive On-policy Self-Distillation (MTA-OPSD), the first token-level adaptive OPSD method. Our method constructs a multi-teacher pool consisting of candidate teachers, and enables every token to adaptively select its preferred teacher during training. Experiments with Qwen3 models on mathematical reasoning tasks show that MTA-OPSD consistently outperforms existing OPSD approaches with an average gain of up to $+5.9$ over the base model, highlighting the importance of adapting teacher supervision at the token level. The code is available at https://github.com/xingzhejun/MTA-OPSD.

★ Nemotron-Labs-Diffusion-Uni: Elastic Self-Refining Unified Diffusion LLM with Cross-Paradigm On-Policy Distillation

Shufan Li, Hanrong Ye, Greg Heinrich, Yonggan Fu, Haotian Zhang, Sifei Liu — Fri, 30 Oct 2026 00:00:00 +0000

95% · New model 95%

We propose Nemotron-Labs-Diffusion-Uni (NLD-Uni), a unified discrete diffusion large language model (DLLM) for visual understanding and generation tasks. Compared with prior unified diffusion models, NLD-Uni introduces two major innovations. First, we propose a unified cross-paradigm on-policy distillation (OPD) framework that successfully transfers the visual understanding ability of autoregressive VLMs and visual generation ability of continuous flow models, to a unified DLLM. Second, NLD-Uni incorporates a novel progressive self-refining diffusion process that gradually corrupts each token across multiple steps during training, as opposed to adopting the conventional masked diffusion formulation that directly corrupts a clean image token to an uninformative mask token. At inference, this design allows NLD-Uni to make intermediate commit to a token, enabling progressive self-refinement for better visual quality. Through extensive experiments, we show that NLD-Uni achieves state-of-the-art performance among unified DLLMs on diverse tasks including visual question-answering, multi-modal reasoning, object grounding, text-to-image generation, and image editing.

Post-Training Large Language Models with LLMON

Fri, 30 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 86%

Large Language Models (LLMs) treat all context tokens uniformly, lacking explicit separation between executable instructions and passive data. This leads to prompt injection vulnerabilities, ambiguous instruction binding, and unreliable execution control. We present a post-training methodology using LLMON (LLM Object Notation), a framework that enables separating instructions from data and requires explicit identifier-based execution. We construct a large-scale structured training corpus of 4.6B tokens spanning general, RAG (Retrieval-Augmented Generation), and tool-calling data, augmented with distractor-infused variants in which competing instructions and irrelevant data segments are present in context but not referenced. Post-training two Granite-4.1 base models (3B and 8B) through supervised fine-tuning followed by reinforcement learning, we show that LLMON-structured supervision reduces attack success rates on indirect prompt-injection benchmarks from 53-100% to 0.02-5.6% while preserving general capabilities. We further introduce attention-based interpretability metrics that diagnose failure modes at individual transformer layers for structured inputs, revealing how our approach reshapes internal attention patterns. Our results demonstrate that explicit execution semantics can be taught through standard post-training without architectural modification.

ATI-OPD: Adaptive Teacher Intervention for Stable Long-Sequence On-Policy Distillation

Jiannan Guan, Youxin Jiang, Yutai Hou, Qiguang Chen, Yirong Zeng, Haitao Wang — Fri, 30 Oct 2026 00:00:00 +0000

94% · New model 94%

On-policy distillation (OPD) trains a student on its own responses using token-level feedback from a teacher. As responses grow longer, the student's prefix can depart from the teacher's own reasoning, weakening its feedback on later tokens. Updates based on this feedback can reinforce the same deviations in later responses and destabilize training. Prior approaches such as length control and loss reweighting limit learning from problematic responses but leave their prefixes unchanged. Fixed cutoffs can also discard later reasoning. We propose \textbf{Adaptive Teacher Intervention for On-Policy Distillation (ATI-OPD)} to improve subsequent teacher guidance by revising the student's reasoning prefix. ATI-OPD replaces the earliest high-disagreement segment with a short teacher segment, then lets the student resume its own reasoning from the revised text. To handle renewed disagreement, it excludes persistently divergent suffixes from the distillation loss while retaining useful student continuation. Training diagnostics show less teacher-student disagreement, fewer gradient spikes, and fewer extremely long responses. Across five mathematical reasoning benchmarks and two teacher-student pairs, ATI-OPD improves average student-only accuracy over standard OPD.

Statutory Article Retrieval from Arabic Judicial Narratives: Family Law and Law of Evidence in Saudi Arabia

Zarah Shibli, Areej Alhothali, Lulwah Alharigy — Fri, 30 Oct 2026 00:00:00 +0000

94% · Evaluation 94%

Reliable legal information retrieval must ground legal reasoning in authoritative statutory sources. In Arabic judicial settings, long factual narratives must be mapped to concise articles that express abstract legal rules, creating a narrative-to-legal-rule abstraction gap. Using a benchmark of 200 synthetic Saudi Arabic judicial cases and 381 articles from the Family Law and the Law of Evidence, we compare word-level and char-4 BM25, zero-shot dense encoders, per-law retrieval, dense-sparse reciprocal rank fusion (RRF), and a proposed Decompose-and-Refine (DaR) pipeline. In DaR, a large language model splits each narrative into disputed legal issues and generates, for each issue, three retrieval channels: a statute-style paraphrase, legal keywords, and a candidate taxonomy chapter. On the whole corpus, the instructed Qwen3-Embedding-8B dense encoder reaches an MRR@10 of 0.849, the best result of any non-decomposed system. Under per-law scoring, this advantage reflects Family Law only: restricted to Evidence articles, the same encoder drops to an MRR@10 of 0.087, and fusion with BM25 raises it only to 0.207. DaR changes this picture, raising Evidence MRR@10 to 0.696 and Family MRR@10 to 0.864. Without decomposition, the two laws favor different strategies; under DaR, the same retriever achieves the best MRR@10 on both.

On-Policy Self-Distillation with Teacher Evolution

Shufan Shen, Zhongni Hou, Chao Bi, Yufei Zhang, Guojun Yin, Shuhui Wang — Fri, 30 Oct 2026 00:00:00 +0000

94% · New model 94%

On-Policy Self-Distillation (OPSD), which aligns a student model's token-level distribution with that of its self-teacher prompted by privileged context, has emerged as a promising direction to mitigate reward sparsity in sequence-level post-training paradigms. However, OPSD often fails to achieve continuous performance gains during training. In this paper, we find that the model's limited self-teaching capability underlies this phenomenon by generating unreliable guidance on certain samples. To address this problem, we propose Teacher-Evolution Self-Distillation (TESD), which enables the self-teacher to generate rollouts and receive environmental feedback based on task correctness and context independence. First, using these rewards as training objectives, we perform teacher-oriented optimization to continuously enhance the model's self-teaching capability during distillation. Second, by estimating teacher guidance reliability from these rewards, we apply adaptive student-oriented optimization to down-weight samples with unreliable guidance. Extensive experiments across multiple datasets and model scales show that TESD consistently outperforms representative token-level and sequence-level methods, highlighting the necessity of improving supervision quality in OPSD.

Adversarial Self-Distillation

Mohammad Sadegh Akhondzadeh, Soroush H. Zargarbashi, Aleksandar Bojchevski — Fri, 30 Oct 2026 00:00:00 +0000

94% · New model 94%

Even though current safety-aligned language models are able to refuse harmful requests, their guard can still be bypassed by adversarial perturbations. As in other domains, adversarial training is considered a standard defense. However, due to the autoregressive nature of language generation, supervised finetuning on template refusals does not necessarily teach the model to refuse. Evidence of this is that the adversarially trained model still complies when the attacker injects a compliant prefill. We propose Adversarial Self-Distillation (ASD), which extends a model's own safe behavior to the space of adversarial perturbations. The teacher (the model itself given the clean prompt) can refuse the harmful request by default, or with the help of an additional instruction in its context. ASD distills this knowledge to the student (which is the same model under attack) at the prefixes that the student visits (due to on-policy sampling). These potentially include compliant openings. ASD achieves lower attack success than state-of-the-art defenses with little loss in multiple-choice accuracy, and it becomes able to recover a refusal even after a compliant opening.

A Negotiation-Aware Hybrid LLM Framework for Intent Interpretation and Service Discovery

Wided Meflah, Hayet Brabra, Walid Gaaloul — Fri, 30 Oct 2026 00:00:00 +0000

94% · Hallucination 94%, Evaluation 56%

Increasingly, users express their needs as natural-language intents rather than structured requests. These intents may be incomplete, high-level, or internally conflicting. A conflicting intent either violates fundamental properties of the cloud paradigm or states mutually inconsistent requirements. Classical composition approaches are designed for structured, low-level, and complete requests. They cannot interpret such intents, discover suitable services from provider catalogs, or negotiate conflicting requirements with users. To address this gap, we propose a negotiation-aware framework for intent interpretation and service discovery. The framework first refines the natural-language intent and dynamically detects conflicts. It then structures the intent into a provider-independent TOSCA template. For each specification, it retrieves candidate service offers through deterministic matching grounded in cloud provider offerings. Some conflicts arise during interpretation. Other requirements prove unsatisfiable during discovery. Both are resolved through iterative natural-language negotiation with the user. We evaluate the framework on 70 diverse intents against single-shot and enhanced-prompting baselines. It improves recall and F1 for both conflict detection and service discovery. It also eliminates hallucinated offers and resolves conflicts at a cost proportional to their number.

SCARL: Sequence Confidence-Aligned Reinforcement Learning

Zhenghong Ke, Jingbo Zhou, Hua Wu — Fri, 30 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 32%

Reinforcement learning with verifiable rewards (RLVR) improves language-model reasoning by scoring sampled responses with a rule-based verifier and reinforcing those that pass. Where a classifier's training couples predicted probability to labels, nothing ties an RLVR policy's confidence in a response to its reward: aligned training would drive responses with the same advantage toward equal confidence, yet the advantage is blind to confidence. Under this sequence-confidence misalignment, a success the policy barely found earns the same update as one it produces with near certainty, and a confident error is corrected no harder than an exploratory one. We propose SCARL (Sequence Confidence-Aligned Reinforcement Learning), which reshapes, but never re-signs, each advantage with the alignment gap between the response's group-relative confidence and the outcome its advantage certifies, normalized into shares of each class's advantage budget. Misaligned samples absorb the force, responses with the same advantage are driven toward equal confidence, and repairing the misalignment doubles as a remedy for the entropy collapse that erodes exploration. SCARL provably reduces to its base estimator when a class's confidences agree, adds no extra passes and one hyperparameter. On Qwen3-4B-Base it lifts average Pass@32 by 2.4 points over DAPO and is best among nine compute-matched methods on both coverage and per-draw accuracy; applied unchanged across four estimators and four backbones, it improves 14 of 16 settings while preserving policy entropy and widening the confidence-correctness separation. Our code is available at https://github.com/kezhh163/scarl.

PIVOT: Progressive Intervention via Online Teacher Correction

Zhihao Yang, Boyuan Zheng, Puhui Tan, Yukun Chen, Kuan Zhang, Longze Chen — Fri, 30 Oct 2026 00:00:00 +0000

94% · New model 94%

Reinforcement learning from verifiable rewards (RLVR) is limited by sparse outcome-level feedback on hard reasoning tasks: when all rollouts in a group fail, rewards provide no learning signal. Stronger teachers can provide additional guidance, but this raises a central question: how can such signals be incorporated into RLVR to support successful rollouts while remaining learnable for the student? We distinguish two complementary mechanisms: trajectory-level intervention, which improves access to reward-bearing trajectories, and token-level supervision, which provides dense targets to guide student learning. Our controlled experiments show that either mechanism alone can remain insufficient, motivating their combination. We therefore introduce PIVOT, which inserts teacher-generated corrections into student rollouts and applies token-level KL supervision to these segments alongside the RLVR objective. Beyond their combination, we systematically analyze where to intervene and how to supervise, yielding a recipe for effective guidance. Across hard logic games and mathematical reasoning benchmarks, PIVOT consistently outperforms existing methods, nearly doubling mean avg@8 on logic games from 19.5% for the best-performing baseline to 35.6%.

Look Only When Needed: Certificate-Guided Visual Routing for Efficient Diffusion VLMs

Donghyeon Cho, Joowon Kim, Seungho Shin, Eunho Yang — Fri, 30 Oct 2026 00:00:00 +0000

94% · Efficiency 94%, Evaluation 39%

Diffusion vision-language models (DVLMs) process their entire input at every denoising step. Since visual tokens make up 89-97% of this input on the benchmarks we study, most of the decoding computation is spent repeatedly processing the same image. Visual token pruning methods partially relieve this burden by removing part of the visual tokens, but they lose the removed information and still process the remaining image at every step. To measure how often the image is actually needed, we compare LLaDA-V's predictions with and without the image at every decoding step and find that removing the image changes a token being unmasked at only 22.8% of steps on average and at as few as 7.9%. We therefore propose LOWN, a decoding method that decides at every step whether to process the image and how many tokens to unmask. LOWN predicts the masked tokens without the image and processes the image only when this prediction disagrees with the last image-conditioned prediction. Agreement certifies the earlier prediction for reuse, and the number of certified tokens sets how many are unmasked in that step. We further observe that the need for the image differs across benchmarks, so LOWN's adaptive decision is needed rather than a fixed schedule. Without training or a tuned threshold, LOWN retains 99.6% of LLaDA-V's score on ten benchmarks at 35.3% of the decoding FLOPs on average. LOWN also generalizes across LLaDA-based DVLMs, composes orthogonally with visual token pruning, and runs under key-value caching for further efficiency gains. Code and project page are available at https://hyeon9698.github.io/LOWN/

FAD-KV: Adaptive Dictionary KV Caching for Efficient Reasoning Decoding

Fri, 30 Oct 2026 00:00:00 +0000

93% · Efficiency 93%

Key-value (KV) caching is a core mechanism in large language models, where intermediate key and value representations are stored to accelerate autoregressive decoding. As generation length increases, KV cache memory and computation grow linearly, making compression critical for efficient long-output reasoning. Dictionary-based sparse KV compression is appealing because it represents each token's KV states as a sparse combination of learned basis elements (atoms) in a dictionary, preserving token-level structure without enforcing a global low-rank constraint. However, existing dictionary-based methods face two practical limitations during decoding: iterative sparse coding methods such as Orthogonal Matching Pursuit add decoding-time overhead, and static dictionaries may mismatch evolving KV distributions under test-time generation. While adaptive dictionary updates can reduce this mismatch, naive append-only strategies lead to dictionary growth during generation. We propose FAD-KV, a \textbf{F}ast \textbf{A}daptive \textbf{D}ictionary KV caching framework for efficient reasoning decoding. FAD-KV integrates (i) a screening-based sparse coding scheme, where a one-shot Top-$n$ selection identifies a small candidate pool followed by lightweight coefficient refinement, and (ii) a fixed-size dictionary adaptation mechanism that reuses inactive atoms instead of expanding the dictionary. By controlling both sparse coding cost and adaptive dictionary growth, FAD-KV provides a memory-bounded and computation-efficient KV compression framework for long reasoning workloads. Experiments on reasoning and long-context benchmarks show that FAD-KV achieves a strong quality-efficiency trade-off while avoiding unbounded dictionary growth during decoding.

MASS: Multi-mask Selective Self-Distillation for Leakage-Resilient Reasoning

Xiuqi Liu, Oveys Delafrooz Noroozi, Upamanyu Madhow — Fri, 30 Oct 2026 00:00:00 +0000

93% · New model 93%

On-policy self-distillation enables the same model to act as student and teacher, with dense token-level supervision obtained by the teacher being given access to privileged information. However, conditioning the teacher on a single reference can concentrate its predictions on one reasoning trajectory and limit transfer to the student. We propose Multi-Mask Selective Self-Distillation (MASS), which approximates the unavailable marginal over diverse reference realizations using multiple independently masked views of one verified solution. MASS aggregates these complementary teacher predictions, selectively transfers reliable token-level corrections, and adaptively focuses masking on influential reference spans. Across mathematical reasoning benchmarks and model families, MASS consistently outperforms strong self-distillation baselines.

Rubricon: Automated Rubric Generation and Refinement for LLM Evaluation and Training

Ulyana Tkachenko, Hui Wen Goh, Guram Gogia, Jonas Mueller — Fri, 30 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, New model 54%

Rubrics enable reward assignment on tasks without verifiable answers, but are expensive to manually produce from scratch. We present Rubricon, a pipeline that generates discriminative rubrics without requiring human preference labels, rubric annotation, or generator-model training. Using any LLM, Rubricon samples multiple rubrics, selects the most discriminative candidate, adds criteria to capture overlooked requirements or response flaws, and prunes criteria that are misaligned or shallow. Alongside LLM self-critique and test-time scaling, Rubricon uses observed criterion pass rates on example model responses to guide refinement. Across five professional, consumer, clinical, and general-use LLM benchmarks, Rubricon produces rubrics whose scores of frontier LLM responses exhibit 9% greater alignment with labor-intensive expert-authored rubrics, compared to the best alternative synthetic rubric generation method among RRD, QSRG, OpenRubrics, and agentic generation via harnesses like Codex. On HealthBench, post-training a Qwen3.5-9B model with Rubricon rubrics yields 4.5% better performance than training with rubrics from the strongest alternative synthetic rubric generator, RRD.

Leaving, Staying, or Living There? Duration Commonsense Reasoning with Human-LLM Co-annotations

Chenyang Yu, Zhaomin Xiao, Yan Huang, Eduardo Blanco — Fri, 30 Oct 2026 00:00:00 +0000

93% · Evaluation 93%

Gardening in the backyard implies living there, while checking into an Airbnb more likely means staying for a few days. Such duration judgments are important for understanding human mobility, travel behavior, and social context. However, duration is rarely stated explicitly in social media posts and must instead be inferred from noisy textual, visual, and contextual cues. In this paper, we investigate whether vision-language models possess such duration commonsense. We create a new corpus of geotagged tweets co-annotated by humans and LLMs to serve as a testbed for this task. We find that state-of-the-art models struggle with duration inference, even when guided by multi-turn Chain-of-Thought prompting. Error analyses further reveal that models often fail to integrate critical contextual information, or they overly rely on prominent visual details while overlooking subtle hashtags and semantic nuances. These findings underscore the significant challenges remaining in context-aware multimodal understanding.

Causal Attention on CPU: A Cache-Aware and Load-Balanced Scheduling Kernel for Prefill

Kwun Fung Lau, Frost Mitchell, Kanya Mo, Emani Ashok — Fri, 30 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 31%

Agentic AI drives significant productivity gains across diverse domains, but concurrent long-context prefills from multiple agents can saturate GPU compute resources. Although CPUs offer a viable alternative, existing CPU attention kernels fall short in adapting to multi-tier cache hierarchies and lack effective thread scheduling. To address this, we introduce CACPU, a high-performance attention kernel featuring tiered cache blocking and an in-kernel task dispatching mechanism designed for causal full-sequence prefilling. CACPU achieves up to a 2.01x speedup over PyTorch SDPA, FlexAttention, and SGLang attention kernels. Integrated into SGLang, it reduces Time to First Token (TTFT) of Llama-3.1-8B by up to 1.74x while preserving model accuracy.

Grounding Multimodal Reasoning with Evidence-Ablated Negatives

Chenxi Liu, Tianyi Xiong, Yanshuo Chen, Heng Huang — Fri, 30 Oct 2026 00:00:00 +0000

93% · New model 93%, Hallucination 58%, Evaluation 31%

Reinforcement learning with verifiable rewards (RLVR) has become a powerful tool for improving multimodal reasoning. However, standard outcome supervision suffers from credit assignment ambiguity, as a model may receive positive rewards by exploiting language priors or dataset shortcuts rather than grounding its reasoning in the requisite visual evidence. This makes high-quality negative responses critical, since effective negatives should remain plausible under the current policy while exposing the shortcut that should be suppressed. In this paper, we propose EVAN, a multimodal reinforcement learning framework utilizing EVidence-Ablated Negatives to train LMMs against their own language-prior failures. EVAN first identifies highly visually dependent examples using a scaled image information gain metric and preserves data diversity with a mixed-ranking strategy. It then constructs evidence-ablated images by masking the most discriminative image patches, eliciting plausible but visually ungrounded responses from the current policy. During policy optimization, these negative responses are incorporated as bounded auxiliary signals under the clean image, penalizing shortcut learning while maintaining RL stability. Experiments across challenging multimodal reasoning benchmarks demonstrate that EVAN mitigates modality imbalance and improves over strong RLVR baselines. Our code is available at https://github.com/DawnLIU35/EVAN.git.

Raincheck: Deferred Multimodal Prefill for Efficient Multi-Turn MLLM Inference

Wenjie Huang, Jun Zhang, Hongyaoxing Gu, Haoyang Huang, Tianqi Xu, Xinyue Hao — Fri, 30 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 34%

Multi-turn conversations with multimodal large language models (MLLMs) reuse input context, but long multimodal inputs increase time to first token (TTFT). Query-aware hard pruning reduces prefill tokens but permanently excludes information that later questions may require. We propose Raincheck, a framework for deferred multimodal prefill that decouples initial token selection from later information availability. An existing pruning method selects tokens for initial prefill; Raincheck preserves and progressively prefills unselected tokens in source order using unused compute capacity during decode. Unfinished prefill continues after answer return during available scheduling gaps. Attention visibility control and affine state mapping enable context updates after generation begins in full-attention and hybrid recurrent models. Prefill-decode fusion reduces repeated weight reads. We evaluate Qwen3-VL-8B and Qwen3.5-9B with 30% initial multimodal-token retention. Across both models and four pruning methods, Raincheck recovers on average 77.7% and 60.4% of the quality lost to matched hard pruning on MultiVerse and ConvBench, respectively. On MultiVerse, the SGLang implementation reduces mean first-turn TTFT by up to 24.1% relative to full prefill, with less than 0.5% increase in mean time per output token (TPOT).

Your Benchmark Is an Empirical Measure Over Difficulty

Yifan Sun, Naicheng Yu, Jiawen Gong, Jingyan Shen, Huan Zhang — Fri, 30 Oct 2026 00:00:00 +0000

93% · Evaluation 93%

A benchmark is not merely a collection of test questions; it naturally induces an empirical measure over difficulty. In this work, we develop this measure-based view of benchmark evaluation. This perspective allows us to identify and formalize two central failure modes: measure mismatch, where the benchmark difficulty distribution differs from the intended difficulty distribution, and coverage gaps, where uncovered regions along the difficulty axis prevent benchmarks from detecting meaningful localized training progress. We propose two diagnostic metrics that capture these two failure modes and provide rigorous theoretical guarantees. Empirically, we analyze 25 popular benchmarks spanning 7 domains using carefully estimated item difficulties. Our results show that many benchmarks implicitly overweight certain difficulty regions, so their leaderboard conclusions should be interpreted with caution; others leave substantial portions of the difficulty axis uncovered, limiting their ability to monitor the model improvement process. Finally, we turn diagnosis into action by introducing BenchFill, a simple difficulty-targeted item rewriting algorithm that effectively reduces local coverage gaps.

Prompt-agnostic Language Models

Frederic Sadrieh — Fri, 30 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 54%

Despite their strong performance, large language models remain highly sensitive to prompt formulation. This issue is typically addressed at inference-time through manual prompt optimization, which introduces overhead and lacks scalability. Conversely, we focus on train-time approaches that build prompt robustness into the model during instruction tuning. A common strategy is to mix multiple paraphrases of each prompt into training. We demonstrate that across trainable parameters, the per-paraphrasing gradients interfere, with 57% exhibiting sign conflicts and the naive gradient average discarding approximately a fifth of the magnitude. Therefore, we develop two gradient-merging methods, Prompt-TIES and Prompt-DARE, that address this interference by leveraging interference-resolution operators from model and gradient merging and applying them to prompt gradients. In this study, we compare the efficacy of our methods to increase model robustness to train-time robustness methods spanning all three established classes: data construction, consistency regularization, and contrastive alignment. The evaluation is conducted across four models from two families at two scales. We observe the interaction between prompt phrasing and the specific examples accounts for 54% of per-example variability on average, and exceeds 70% on tasks such as ANLI and WiC. The gradient-merging methods directly improve this local robustness and nearly halve per-example variance on fully fine-tuned models at little cost to overall accuracy. The other train-time methods enhance raw performance, but fail to improve per-example variance reliably. Code: https://github.com/FSadrieh/prompt-agnostic-llms. Model checkpoints: https://huggingface.co/prompt-agnostic-language-models.

From Natural Language to Real Schematics: An Automated Workflow for Power Electronics Schematic Synthesis

Bowen Su, Huan Chen, Jiahua Ying, Ran Ou, Weilong Chen, Li Xue — Fri, 30 Oct 2026 00:00:00 +0000

92% · Hallucination 92%, Agents 80%, Evaluation 61%

Designing power electronics converters traditionally requires intensive human expertise across topology sizing, commercial component selection, circuit simulation verification, and schematic drafting. While Large Language Models (LLMs) show promise in electronic design automation (EDA), applying standalone LLMs directly to board-level power electronics schematic synthesis suffers from severe engineering hallucinations, including inaccurate analytical sizing, fabricated or mismatched component part numbers, violated voltage derating margins, and broken topological connectivity. In this paper, we propose a Bounded Dual-Loop Automated Schematic Synthesis Framework that translates natural language specifications into verified, editable KiCad schematics and manufacturing deliverables. Our workflow integrates six stages across three functional columns: (1) requirement parsing and knowledge-driven power stage synthesis covering non-isolated (Buck, Boost) and isolated/resonant (Flyback, DAB, half-bridge LLC) topologies; (2) contract- and evidence-coordinated component selection grounded in distributor APIs and OEM datasheets, coupled with closed-loop ngspice transient verification and stress feedback; and (3) modular control hardware integration and multi-format deliverable generation featuring native KiCad netlist round-trip verification. End-to-end case studies on a 24-36V to 12V,2A Buck converter (29 positions, 83 verified terminals) and an isolated 48V to 24V, 0.5A, Dual-Active Bridge (DAB) converter (87 positions, 341 verified terminals), alongside comparative evaluations against a Vanilla LLM baseline, demonstrate that our framework improves component verification coverage from 22.2 to 98.6, nominal value consistency from 0 to 97.7, and SPICE simulation pass rates from 0 to 100.

Keepace.ai: A Technical Report on a Large Language Model for Sports and Health

ai Team, Bowen Wu, Heng Wang, Jufeng Mi, Kang Liu, Kun Li — Fri, 30 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 84%, Hallucination 51%

Large language models hold promise for delivering safe, scientifically grounded, and personalized exercise guidance at scale, yet the sports-and-health domain still lacks curated training corpora and rigorous evaluation benchmarks. Keepace.ai is a domain-specialized LLM built on Qwen3.5-27B through post-training, currently optimized for Chinese-language exercise guidance across three tasks: workout plan generation, sports-and-health knowledge QA, and personal sports data analysis. In internal rubric-based evaluations designed with domain experts, Keepace.ai outperforms its base model and larger general-purpose baselines, with the most pronounced improvements in exercise safety, boundary compliance, structured output completeness, and scientific accuracy. This report summarizes the current design, data pipeline, safety-first evaluation framework, and evaluation results of Keepace.ai.

ParaTTT: Parallelizing Test-Time Training Layers with Deep Memories

Fri, 30 Oct 2026 00:00:00 +0000

92% · Efficiency 92%, New model 90%

Test-time training (TTT) layers replace the linear fast-weight memory of modern recurrent architectures with a small neural network trained online by gradient descent. This nonlinear memory is more expressive than the linear memories of Gated DeltaNet and Mamba-2, but is difficult to parallelize: existing implementations such as TTT-MLP rely on sequential, block-wise weight updates that run several times slower than chunkwise-parallel linear recurrences. Moreover, their block-wise execution only approximates the token-by-token online recurrence, whereas linear recurrences admit an exact chunkwise-parallel evaluation. We introduce ParaTTT (PTTT), an algorithm for parallelizing arbitrarily deep nonlinear memories by iteratively refining estimates of the inner-loop gradient trajectory. Starting from a guess of the hidden-layer gradients, each refinement consists of three chunkwise-parallel passes: a forward pass that evaluates the resulting feature trajectory, an exact gated-delta-rule pass that computes the readout trajectory and its feature gradients, and a backward pass that propagates these gradients into a new estimate. The exact online gradient descent recurrence is a fixed point of the iteration, while decoding requires no refinement iterations. We prove geometric convergence for a two-layer variant under sufficient contraction conditions. In practice, two iterations suffice. With two refinement steps at 4K context, PTTT evaluates the inner-loop trajectory $32\times$ faster than TTT-MLP with the block approximation at matched error, while the full 340M model trains $4.9\times$ faster. Models trained at 340M and 1.3B parameters match or exceed Gated DeltaNet and Mamba-2 and remain competitive with same-size Transformers on language-modeling perplexity and zero-shot accuracy, with better recall than the linear recurrences.

Value Probes: Learned Value Functions for Beam-Search Red Teaming

Jonathan F. Jao, Andrew Kyle Lampinen — Fri, 30 Oct 2026 00:00:00 +0000

92% · Evaluation 92%, Efficiency 65%

Automated red-team auditing of language models is an important way to evaluate safety-relevant behaviors. These audits implicitly search a tree of possible conversation pathways: at each turn an auditor model decides which conversation to continue, and an LLM judge scores the dialogue. We emphasize that this auditing procedure is a Markov decision process, where states are the conversation history, actions are candidate responses, and rewards are judge scores for the target behavior. By adopting this perspective, we introduce two major improvements to auditing. First, we propose adopting beam search for the audit rather than relying on the auditor alone. We then interpret the problem of finding states that are likely to lead to problematic behavior as value estimation. Surprisingly, we show that this value function can be estimated by linear probing from the target model's activations to predict discounted future rewards - a method we term "Value Probes." By using Value Probes as a heuristic in our beam-search audit, we are able to substantially improve auditing success. Across 18 target behaviors and 3 scales of models, our approach raises peak audit scores by roughly 50% compared to (token-matched) standard auditing approaches. Furthermore, we show that Value Probes can be fit using off-policy data from smaller models. We also ablate individual components, and show that our method offers a comparably-performant and more efficient alternative to more expensive language model follow-ups. We hope these methods will improve language model auditing and lead to safer models.

Co-Evolving Policy and Rubric via Capability Awareness for Open-Ended Generation

Fri, 30 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 37%

Rubric-based reinforcement learning (RL) offers interpretable, fine-grained supervision for aligning large language models (LLMs) on open-ended generation tasks. However, static rubric criteria may become saturated or remain beyond the policy's exploration, rendering the reward signals insufficiently discriminative. While recent efforts seek to guide sampling exploration or evolve rubrics, they fail to explicitly factor in how rubric difficulty should track the evolving policy's capabilities, resulting in a persistent mismatch between the optimization objective and the policy's current capability frontier. In this paper, we propose \textbf{Co}-Evolving Policy and Rubric via \textbf{C}apability \textbf{A}wareness (\textbf{CoCA}), a closed-loop framework that jointly adapts the rubric and policy through criterion-level diagnosis. Specifically, CoCA contrasts query-only and rubric-guided rollouts to characterize each criterion relative to the current policy, recalibrates criteria with mismatched difficulty, and identifies guidance-revealed capability gaps for frontier-based learning. Based on this insight, CoCA either expands rubric-conditioned frontiers through guided optimization or transfers qualified gains through on-policy self-distillation. As the policy updates, CoCA improves the capability diagnosis and enables more reliable distillation, thereby completing the co-evolution loop. Extensive experiments across diverse benchmarks and LLMs demonstrate that CoCA consistently improves performance over rubric-based RL methods. Moreover, CoCA promotes the internalization of guidance-revealed capabilities into query-only behavior while mitigating capability degradation.

Beyond fabrication: Misattributed numbers in papers written by AI research agents

Mosab Alfaqeeh, Omar Hourani, Muath Hourani, Hashem Alfaqeeh — Fri, 30 Oct 2026 00:00:00 +0000

92% · Evaluation 92%, Agents 80%, Hallucination 76%

Verifying numerical claims in research written by AI agents is an important but challenging problem. Recent benchmarks show that many agent-written papers fail integrity checks, but such verdicts, often produced by a single LLM judge, identify failure without explaining its nature, and existing checks either test a paper for internal consistency or test whether a reported value appears somewhere in an agent's output. We re-analyze all 540 released judgments from a recent public benchmark of AI research agents (nine agents, fifteen open research problems) and manually label 256 numerical claims from 18 submissions. Failures concentrate on alignment between the paper and the underlying work rather than overt cheating, and about four in five unsupported claims correspond to values in the agent's own files that come from the wrong experiment, metric, or run. We introduce a deterministic claim-to-evidence checker that scores whether the surrounding context identifies the claimed result and calibrates each claim against numerical decoys. With its rules fixed before labeling, it certifies claims with over 90% precision against blinded annotators (78 of 85; Cohen's kappa = 0.89), while verifying only a subset of claims; most claims it cannot verify are nevertheless traceable, so failure to verify is not evidence of falsehood. An LLM verifier given the same evidence recovers more traceable claims but accepts about three times as many misattributed ones, so the two are best used together. Applied to 99 submissions, the checker reveals large differences in traceability across agents, and a second judge from a different vendor preserves the official agent ranking while passing twice as many papers. These findings motivate provenance-first auditing of AI-produced research.

When Does Closed-Loop Curriculum Adaptation Help GRPO? An Analysis on Countdown

Yuv Raj Pant, Moshiur Rahman Tonmoy, Kaiqun Fu — Fri, 30 Oct 2026 00:00:00 +0000

92% · New model 92%

Reinforcement fine-tuning has become an effective post-training paradigm for improving the reasoning ability of large language models, but it remains sample-inefficient when the training distribution is mismatched to the model's current capability. In group-relative objectives such as Group Relative Policy Optimization (GRPO), problems that are too hard often produce uniformly failed rollouts, while problems that are too easy provide little useful advantage signal. In this work, we introduce CLOSEDLOOPGAUSSIAN, a lightweight adaptive curriculum method for efficient reinforcement fine-tuning on ordered reasoning tasks. Instead of following a fixed easy-to-hard schedule, CLOSEDLOOPGAUSSIAN preserves the structure of an Easy-to-Hard Gaussian sampler while dynamically adjusting its pace and spread using periodic probe estimates of per-difficulty accuracy. The controller accelerates training when the model is ahead of a reference curriculum, slows down when the model falls behind, and broadens sampling when earlier difficulties regress. This feedback mechanism keeps training focused on tasks that are challenging but still learnable, without changing the reward function, model architecture, or underlying GRPO optimizer. Experiments on symbolic arithmetic reasoning task show that CLOSEDLOOPGAUSSIAN improves exact-match accuracy over vanilla GRPO and fixed open-loop curricula, with the strongest gains in intermediate-difficulty regimes where group-relative learning signals are most informative. Codes are available at https://github.com/yuvrajpant56/Closed_Loop_Curriculum.

Revisiting Over-refusal in Safety-aligned Large Language Models: A Representation and Optimization Perspective

Chao Chen, Pengyang Shao, Wei Qin, Shijie Hao, Fei Shen, Richang Hong — Fri, 30 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 32%

Over-refusal is a critical side effect of safety alignment in large language models (LLMs), causing harmless but superficially risky queries to be mistakenly rejected. Despite the effectiveness of existing mitigation methods, why safety-aligned LLMs may still reject harmless queries under correct safety supervision remains insufficiently understood. To address this gap, we analyze over-refusal through the lens of representation and optimization. We make two observations: 1) the representations of harmful and over-refusal samples are visually distinguishable at certain intermediate layers but increasingly overlap in deeper layers; and 2) harmful-sample gradients exhibit high intra-class similarity and concentrate in a low-rank subspace. Moreover, they are more strongly aligned with refusal-target gradients than with correct helpful-target gradients on over-refusal prompts. This imbalanced gradient coupling can bias over-refusal samples toward refusal, while the reduced apparent representational separation in deeper layers may make it more difficult for the model to maintain a clear boundary between harmful and over-refusal samples. To this end, we propose GOAL (Gradient Orthogonalization with contrAstive Learning), a safety-alignment framework for mitigating over-refusal. Specifically, GOAL consists of two components: 1) to prevent imbalanced gradient coupling from biasing the optimization of over-refusal samples toward refusal, we orthogonalize over-refusal gradients against the low-rank refusal subspace extracted from harmful gradients; and 2) to explicitly enhance the separation between harmful and over-refusal representations, we introduce a contrastive learning objective. Extensive experiments across multiple models and benchmarks demonstrate that GOAL effectively mitigates over-refusal while preserving safety performance and general capabilities.

Turing LLM Post-Training: Calibrated Distillation and Efficient Dynamic-MoE Execution

Yue He, Jiahui Hu, Shaman Tang, Hua Zhou, Qiman Wu, Menglin Li — Fri, 30 Oct 2026 00:00:00 +0000

92% · New model 92%, Efficiency 78%, Evaluation 64%

We study post-training for dynamically routed mixture-of-experts (MoE) models, focusing on integrating independently optimized specialists while keeping consolidation efficient. Our approach combines calibrated multi-teacher on-policy distillation (C-MOPD), which matches each selected teacher on a local Top-$K$ support with teacher-wise loss normalization and clipping, with an execution path that uses post-capacity bounded-width dispatch, final-route replay, and asynchronous rollout--training overlap. On Turing-20B-A2B, the selected model achieves a four-domain normalized integration score of 0.9988 and retains at least 90.16\% of the measured specialist gain in every domain. At matched $K=32$, the full calibration objective scores 0.9962, compared with 0.9724 without calibration and 0.9834 without clipping. At a matched 8K response cap, the model exceeds both shared SFT and Qwen3-8B on all four reasoning-enabled mathematics and code benchmarks (unweighted mean: 85.92, versus 80.33 for Qwen3-8B and 74.21 for SFT), and improves over SFT on ten of twelve broader evaluation metrics. In a PPU-based profile, the combined execution configuration reduces consolidation step time from 398.7 to 139.3 seconds, an observed 2.86$\times$ speedup over serial execution. These results demonstrate strong specialist integration alongside more efficient consolidation, while reflecting the complete training and execution framework rather than any single component.

Locating and Closing Information Gaps in Multi-Agent Systems

Yaxin Gong, Chongming Gao, Gangyi Zhang, Chenxiao Fan, Leyang Shen, Jiakai Wang — Fri, 30 Oct 2026 00:00:00 +0000

92% · Agents 92%, Evaluation 44%, New model 34%

Multi-step Web tasks often require several LLM agents to gather and combine evidence. The communication graph of a fixed workflow restricts which agents can read each other's outputs: evidence retrieved by one agent may never reach the agent that needs it, leaving an information gap and an incomplete or incorrect answer. Adding a connection can deliver the missing evidence, but can also lead the receiver to adopt an incorrect upstream conclusion that displaces a correct answer; the final output alone does not separate the two. We propose Selective Probing and Link Insertion for Closing information gaps in multi-agent Execution (SPLICE), which locates information gaps per query and decides what to adopt after adding an edge. After the original workflow runs, SPLICE uses the query and node outputs to locate a possible gap, adds one directed edge from a source to a receiver to deliver the relevant evidence, and re-executes the receiver and affected downstream nodes, producing a candidate answer. The adoption decision reads both the final answers of the two runs and the process changes the edit caused inside the workflow: how far the receiver moved, where its new content came from, and whether the change reached the final answer. A selector trained on these process features chooses between adopting the candidate answer and retaining the original. On five benchmarks, SPLICE raises the task score on Web tool use by 8.0% and the mean score on four reasoning and planning benchmarks by 5.7% relative to the static skeleton, at 1.25 to 1.44$\times$ the skeleton's cost, and transfers zero-shot to four unseen backbones.

Latent Path Exploration via Potential-Guided Subtrajectory Balance

Xiaoling Zeng, Qi Zhang — Fri, 30 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 30%

Continuous latent reasoning enables language models to perform intermediate computation without generating explicit reasoning traces. Advances in stochastic latent reasoning allow multiple paths to be sampled in continuous space, but stochasticity alone does not ensure effective exploration. A key challenge is guiding stochastic transitions between intermediate latent states toward useful reasoning paths. We introduce SubLAT, which couples a relative answer potential with Subtrajectory Balance to learn stochastic latent exploration. The relative potential evaluates the answer support of sampled states against replacement states, with a correctness signal at the terminal state. Subtrajectory Balance relates potential differences to forward and backward transition densities across subtrajectories of different lengths. The model is jointly trained with answer supervision, CoT supervision, and representation distillation. Experiments on arithmetic and deductive reasoning show increasing solution coverage with larger sampling budgets across in-domain and out-of-domain benchmarks, as well as improvements on several structural OOD benchmarks. Trajectory analyses reveal progressive divergence among sampled latent paths and higher mean relative potentials for correct trajectories across arithmetic benchmarks.

Don't Lose the Thread: Empowering Long-Horizon LLM Agents with Cognitive Resource Self-Allocation

Zheng Zhu, Xiaozhuan Liang, Xi Chen, Bei Yu, Jiaya Jia — Fri, 30 Oct 2026 00:00:00 +0000

92% · Efficiency 92%, Agents 86%, New model 55%

Agents powered by large language models (LLMs) have demonstrated remarkable progress in solving complex reasoning tasks. However, LLM agents often falter on long-horizon tasks due to cognitive overload, as their working memory becomes cluttered with expanding and irrelevant information, which dilutes their attention and hinders effective planning and reasoning. To mitigate this challenge, we introduce Cognitive Resource Self-Allocation (CORAL), a novel reasoning paradigm that empowers agents to proactively optimize their context. Implemented as an agent-callable working memory management toolset, CORAL allows an agent to checkpoint its task progress and verified facts, and adaptively initiate a new problem-solving episode by resetting cluttered context while preserving only the checkpointed information, enabling it to resume reasoning from a clean state. We further enhance the agent's checkpointing capabilities using a Multi-episode Agentic Reinforced Policy Optimization algorithm. On the GAIA benchmark, CORAL with an 8B model achieves 42.8% average accuracy, with particularly strong gains on the most challenging long-horizon tasks (Level 2 and Level 3). Crucially, these gains come at a substantially lower inference cost: CORAL consumes 6-7$\times$ fewer input tokens than a ReAct agent. It also generalizes to multi-hop Knowledge QA benchmarks and to the considerably harder BrowseComp, where it improves a strong proprietary agent by 8.0 points.

A Case Study of a Multi-Agent Kernel-Optimization Loop That Learns a New AI Accelerator

Byonggon Chun — Fri, 30 Oct 2026 00:00:00 +0000

91% · Efficiency 91%, Agents 80%, Evaluation 42%

Frontier large language models write correct CUDA kernels for most single-operator benchmark tasks, yet their accuracy collapses on accelerators whose architecture and programming language they have rarely seen. We present a case study of an LLM agent loop built for such accelerators. Our example is the FuriosaAI RNGD NPU, programmed in a recent, sparsely documented Rust DSL, where the loop optimized three Gemma-4-12B decode kernels over 12 days on a shared, scarce device. The loop builds its own target knowledge: it writes and checks a summary of the DSL documentation, measures the chip with microbenchmarks, records compiler-derived legality rules, and revisits this knowledge when the toolchain changes, rechecking rules and measurements where it can. A role-based search over per-kernel Design Trees explores whole-kernel designs and accumulates small accepted changes in the best version of each kernel. Because device time is scarce and readings are noisy, cheap checks screen new designs before any device run, and late in the run a change was accepted only after repeated comparisons with the best kernel on the device. The final kernels reach a 9.293x geomean speedup over the official baseline, each within 1.25x of the cycles needed just to move its weights. Our prompts, profiler and acceptance test are available at https://github.com/bgchun-fs/multi-agent-kernel-gen-loop.

Implicit Minor Detection from Multi-Turn Dialogue in High-Risk AI Interactions: A Constrained Self-Evolving LLM Harness

Xiaohan Zhang, Ying Zhao, Ruilin Guo, Xuanyuan Yu, Kaibo Huang, Yukun Wei — Fri, 30 Oct 2026 00:00:00 +0000

91% · Evaluation 91%

In high-risk text interactions, a system may need to estimate whether the current speaker is likely to be a minor before applying age-conditioned safeguards. Many text-based dialogue systems lack strong identity signals such as account information, verified identity, facial data, or device data, and must rely on weak cues that emerge across multiple turns. We formulate implicit minor detection under weak identity information as a high-risk user-attribute inference task and propose a constrained framework comprising a large language model, a Harness, and a Maker-Checker protocol. The framework controls when inference is invoked, requires structured dialogue-linked evidence, and subjects validation-selected Harness revisions to version-level review. On a controlled Chinese benchmark comprising 8,328 multi-turn dialogues, our method improves average precision and minor-class F1 by 7.60 and 6.37 percentage points, respectively, over direct inference using the same LLM family.

A-RSI: Adaptive Task Refresh for Zero-Data Multimodal Self-Improvement

Yu Liu, Youxing Li, Yuedong Tan, Zhiwei Yang, Zihan Su, Hanqi Feng — Fri, 30 Oct 2026 00:00:00 +0000

91% · New model 91%, Evaluation 54%

Recursive self-improvement requires models not only to acquire useful training experience, but also to decide when their current experience has ceased to be useful. Existing methods often focus on what experience to generate next, while paying less attention to when fresh experience should be introduced. Many also construct training targets through self-consistency or majority voting, making supervision depend on the model's own uncertain predictions. We address these two questions separately: training dynamics determine when to acquire new experience, while the answers to new tasks remain independently checkable. A-RSI monitors reward variation and accuracy in training logs to refresh a task pool when its learning signal declines, without consulting validation or test scores. A static library of model-authored programs produces multimodal questions and derives their reference answers; we specify the task-generation contract, automatically verify chosen properties of the reward and refresh rules, and separately audit the generated answers. The training loop uses no external training examples or human-written answer labels. On the primary eleven-benchmark suite, A-RSI raises the performance average from 57.77 to 61.05 with 800 fresh tasks.

Latent Decision Flow Rewards for Clinical Guideline Reasoning

Gaurav Bhatt, Marinka Zitnik — Fri, 30 Oct 2026 00:00:00 +0000

91% · New model 91%, Evaluation 54%

Clinical practice guidelines encode evidence-based recommendations as conditional decision flows, which makes them a natural source of process supervision for training reasoning models. Two properties of clinical reasoning make this supervision difficult to use. A decision flow consistent with a guideline does not by itself determine the correct answer for a given patient, which we call the path--answer gap, and several valid decision flows can support the same clinical decision, which we call path multiplicity. Current training approaches address parts of this problem. Answer rewards supervise the final decision, and process rewards computed against a reference flow supervise one valid decision flow per case. Supervised fine-tuning on annotated flows teaches guideline-consistent reasoning but constrains the model to one trajectory per case and can reduce answer accuracy. We introduce GUIDER, a method that jointly optimizes a decision flow controller and a reasoning policy. The controller learns to generate plausible decision flows from guideline-derived supervision, and the policy learns to generate a decision flow and a final answer for each patient case. GUIDER computes the process reward for the policy over several controller-generated flows, which accounts for path multiplicity, and gates this reward on answer correctness, which addresses the path--answer gap. Joint training allows the process supervision to evolve with the policy. On MedGUIDE, GUIDER with Qwen3-14B achieves 82.5 weighted accuracy and 73.5 guideline adherence, 3.3 and 3.1 points above the strongest clinical reinforcement learning baseline. With Qwen3-14B and Gemma4-12B backbones, GUIDER also improves over CRPO, MRPO, and FaithMed on the out-of-domain benchmarks AMEGA-LLM, RealMedQA, and CancerGUIDE.

Where to Learn and How Far to Roll Out: Efficient Adaptive Training for GUI Agents

Yuxiang Chai, Quyu Kong, Panrong Tong, Steven Hoi, Yue Wang — Fri, 30 Oct 2026 00:00:00 +0000

91% · New model 91%, Efficiency 74%, Agents 70%

Trajectory-level GRPO is a common approach for training multi-step GUI agents, but repeatedly rolling out complete trajectories from the task root can be inefficient: task-level rewards provide coarse supervision over which intermediate decisions require improvement, while already-mastered prefixes are repeatedly re-executed before reaching difficult states. We introduce a two-stage adaptive training framework that jointly addresses where to learn and how far to roll out. We first construct reusable intermediate checkpoints from successful expert trajectories and select informative pivots using action disagreement, which measures uncertainty directly from sampled GUI actions. Stage I performs one-step RL at these pivots to improve uncertain local decisions. Stage II then progressively expands policy responsibility through adaptive Local, Suffix, and Root training according to task progress. Recoverable checkpoints enable repeated training from intermediate states without replaying the preceding trajectory. Across MobileGym and CUA-Gym-Hub, our method reduces environment interaction by 56-61% relative to Vanilla GRPO while improving task performance, demonstrating that jointly allocating training across states and rollout ranges can substantially improve GUI-agent RL efficiency.

Robustness of Chatbot LLM Judgments of Presidential Actions and Hypothetical Reforms

Oleg Vasilyev — Fri, 30 Oct 2026 00:00:00 +0000

91% · Evaluation 91%, Hallucination 30%

Chatbots are increasingly consulted about public affairs, and their judgments of policies are useful only if they are robust. We measure this robustness in the setting of an ordinary user, through consumer chat interfaces. For 30 days, five chatbot configurations (Mistral, Gemini Flash, Claude Sonnet, and Claude Opus at two effort settings) were asked twice a day to score the long-term effect on the United States of 40 actions of the last four presidential administrations and of 15 hypothetical reforms, each with a confidence and a short comment. The judgments are robust: all configurations agree on the sign and the ranking of the administrations' mean scores, and two different configurations asked on the same day give opposite signs to the same action in about 4% of cases. Responses from different days differ more than same-day responses, but the difference grows slowly over the month, except where the service itself changed. We identify three such changes, which shifted the judgments while leaving little or no trace in the interface. Treating the comments as a retrieval task, we find them specific to each item and consistent across runs and chatbots, with lexical and embedding similarities agreeing. We release the dataset of 16,500 judgments.

Incentivizing Self-Learning in Large Language Models via Reinforcement Learning

Lingkai Kong, Yuzhe Gu, Haiteng Zhao, Zijian Wu, Youqing Fang, Zhouqi Hua — Fri, 30 Oct 2026 00:00:00 +0000

91% · New model 91%, Evaluation 86%, Agents 56%

Large language models (LLMs) draw on parametric knowledge and retrieval-augmented generation (RAG) to solve knowledge-intensive tasks. Yet problem solving in unfamiliar domains calls for tightly coupled knowledge acquisition and reasoning, beyond factual retrieval and aggregation or learning solely from a preassembled context. We study self-learning -- autonomously retrieving, studying, and applying knowledge from long-form reference materials on the fly. To investigate this capability, we formulate a textbook-based mathematical problem-solving task. LLMs must navigate textbook exposition, learn and apply relevant principles, and construct proofs for exercises, with reference solutions withheld. We introduce a BookNavigator interface to support this human-inspired study process and incentivize self-learning through reinforcement learning (RL) with fine-grained rewards. Experiments show that our method improves performance across in-domain and out-of-domain textbooks and subjects, raising Intern-S2-Preview-35B's overall proof accuracy from 51.5% to 79.5% and outperforming RL-trained textbook-free and fixed-retrieval baselines. These gains extend to complex long-context benchmarks involving different task types, with improvements over the base model of 1.7 and 2.3 percentage points on BrowseComp-Plus and CL-bench, respectively. Together, these findings suggest that RL can foster a domain-agnostic self-learning strategy, highlighting the potential of self-learning from long-form corpora to strengthen broader problem-solving capabilities on complex, long-context tasks.

Learning Calibrated Confidence for Multi-Agent Collaboration

Chenxiao Fan, Chongming Gao, Gangyi Zhang, Yaxin Gong, Leyang Shen, Jiakai Wang — Fri, 30 Oct 2026 00:00:00 +0000

91% · Agents 91%, New model 84%, Hallucination 54%

Complex information seeking on the Web often requires collecting evidence from multiple documents and reasoning across them, as in multi-hop question answering. Multi-agent systems address such tasks by dividing work among specialized agents, where upstream agents produce intermediate answers and supporting evidence that downstream agents build upon. However, upstream answers vary in correctness, and most systems provide no quantitative signal about it, leaving the downstream agent to judge whether to trust or verify what it receives based solely on the textual evidence. In this paper, we propose MASCAL. Each agent learns to produce a correctness probability for its answer by jointly optimizing task performance and confidence calibration in reinforcement learning; the upstream agent passes this probability at handoff, giving the downstream agent a quantitative basis for its decisions, and the final answer also carries a calibrated probability, quantifying the reliability of the system output. To guide the downstream agent in using this signal during training, we introduce confidence-aware weighting, which increases the learning weight at positions responsive to the upstream probability, so that the agent learns more from decisions where it actively uses the signal. Across three collaboration architectures spanning Web retrieval and reasoning tasks, MASCAL improves task accuracy and calibration quality for both agents, and generalizes to deeper collaboration chains and different model scales. Intervention analysis confirms that upstream confidence affects downstream decision behavior, and the learned final confidence supports selective answering and targeted rechecking.

Training language models in simulation for psychiatric intake

Fri, 30 Oct 2026 00:00:00 +0000

91% · Evaluation 91%, New model 74%

In psychiatric practice, the clinical intake interview remains a primary diagnostic instrument, requiring the clinician to establish rapport while disentangling complex, overlapping presentations. Here, we show that multi-turn reinforcement learning (RL) in simulation environments improves language model performance on structured text-based psychiatric intake, spanning history-taking, differential diagnosis, and initial management planning. In a randomized, double-blind crossover study ($N = 60$ matched scenario pairs, 120 encounters) comparing the RL-trained agent against its identically prompted base model, 32 board-certified psychiatrists role-played standardized patients while a separate panel of 15 independent, board-certified psychiatrists scored the de-identified transcripts. The trained agent outperformed the base model in 76.7% of matched scenario pairs ($p < 0.0001$), with gains on an adapted clinical-examination rubric ($d_z = 0.77$, $p < 0.0001$), patient-centered communication (82.0% vs. 65.2%, $p < 0.001$), and differential diagnosis appropriateness (68.3% vs. 33.3%, $p < 0.001$). Across 127 psychiatrist-validated simulated intake scenarios, the trained agent nearly doubled its base model's clinical rubric score (0.706 vs. 0.379; composite reward 0.564 vs. 0.294, $p < 0.0001$). To systematically probe conversational safety, we established two Delphi-consensus taxonomies, of clinical safety risks and relational competencies, from which 256 challenging adversarial presentations were derived; we find that training in simulation environments more than halved under-triage of acute safety risks (12.9% vs. 30.5%, $p < 0.0001$). Simulation environments thus provide a controlled testbed for training and stress-testing psychiatric conversational agents, with gains that transferred to psychiatrist-enacted encounters, though residual gaps require safeguards and clinician oversight.

Training Foundation Models with a Single Algorithm

Fri, 30 Oct 2026 00:00:00 +0000

91% · New model 91%

Foundation models are typically trained in stages: pretraining, midtraining, supervised fine-tuning (SFT), and reinforcement learning (RL), with changes in both data and learning algorithms. We show that this algorithmic separation is not fundamental. By deriving suitable RL objectives as maximum-likelihood fitting to model-generated target distributions, we place all of these stages under a single likelihood-based training algorithm, unified training. This perspective turns the foundation-model training lifecycle into a problem of choosing which data to train on, in what proportions, and when, with conventional staged training as a special case. We instantiate this framework in a unified training system that mixes static data and RL generations within a single trainer. Across controlled and real-world experiments, unified training improves the accuracy-compute tradeoff and task coverage while preserving broader capabilities, and provides a stronger starting point for continual learning. This work lays the groundwork for a unified science of foundation-model training, connecting mathematical principles, training infrastructure, and the development of model capabilities.

Detecting Covert Collusion between AI Agents through Partner-Outsider Positive Listening

Kevin Yandoka Denamgana\"i — Fri, 30 Oct 2026 00:00:00 +0000

90% · Evaluation 90%

AI agents that share a secret convention can coordinate through messages that look innocent, so a monitor that reads only the messages can miss them. We propose a behavioural test: a partner that uses a code reacts to the exact wording of a message, whereas an uninvolved agent placed in its seat does not. Partner--Outsider Positive Listening (PoPLi) is the difference between the two agents in how much their action changes when the message is reworded with its meaning kept, and it is computed after the fact by replaying logged prompts. In a simulated market with five conditions (no collusion, a joint reward with neutral messages, the same with an openly stated intention, a covert first-letter code, and the code with a stated intention), PoPLi is between 0.48 and 0.59 whenever the code is present (95% intervals all above 0.29) and between -0.10 and -0.01 without it. Message-only monitors did not separate the conditions: one reading the speaker's messages answered "no" in all 50 markets, and one reading every seller's messages flagged 9 of 10 markets with no collusion. PoPLi detects the use of a private channel, not collusion as such, and it cannot detect collusion through incentives alone or codes carried by meaning.

Two Birds with One Stone: Skill-Augmented Generation via Unified Retrieval and Utilization

Zihan Luo, Jiahui Kang, Wei Zhou, Qilin Fan, Mingyan Li, Chao Chen — Fri, 30 Oct 2026 00:00:00 +0000

90% · New model 90%, Evaluation 42%, Agents 36%

Skills, as reusable external knowledge, have demonstrated substantial potential to enhance the ability of Large Language Models (LLMs) to solve complex tasks. However, existing skill-augmented generation methods typically rely on separate models for skill retrieval and skill utilization, resulting in disconnected optimization of these two processes. To address this gap, we propose Skill-Uni, a unified reinforcement learning framework that jointly optimizes skill retrieval and utilization. Specifically, we adopt a generative skill retrieval paradigm that enables a single model to both learn how to retrieve relevant skills and leverage them for task solving. To effectively train this unified framework, we introduce a two-stage training strategy consisting of warm-up instruction tuning and GRPO, allowing the model to progressively improve its skill retrieval and utilization capabilities. Extensive experiments on five skill-related benchmarks demonstrate the effectiveness of Skill-Uni, yielding average performance gains of 66.59% and 34.82% over the base Qwen3-4B and LLaMA3.1-8B models, respectively. Besides, Skill-Uni is compatible with various skill retrieval engines, demonstrating its flexibility and generalizability. Additional in-depth analyses also provide insights into the effectiveness of Skill-Uni.

Routed Continual Learning

Fri, 30 Oct 2026 00:00:00 +0000

90% · Efficiency 90%, Evaluation 45%, New model 32%

Agents deployed at scale repeatedly encounter related problems, and it is natural to let them learn from each other's experience rather than starting over each time. Unfortunately, there is no single way to retain that experience that works well across settings, and no way to know in advance which future problems a given experience will help with. A global history preserves every detail but becomes expensive and mixes together experience that may not belong together; compressing that history into one fixed representation limits the cost but can throw away valuable details. We introduce Routed Continual Learning (Routed CL), an architecture that decides how to partition and retain experience. Past trajectories are grouped into clusters, each of which maintains state using one memory system from a fixed library. Before each run, an LLM-based router either reuses an existing cluster or creates a new one, basing its decision on the problem statement, memory system descriptions, and cluster summaries. We also introduce Context Memory Adapter, which mediates between the memory system and the agentic loop. We evaluate the architecture on a heterogeneous stream that interleaves 301 instances from the six CL-Bench tasks with 97 banking instances from $\tau^3$-Bench. Across four language models and five matched rollouts, Routed Continual Learning raises normalized reward over a single global in-context learning (ICL) history by 15-30% of the maximum attainable normalized reward, achieves higher mean normalized reward than ICL-only clustering, and reduces estimated API cost by 65-72% relative to Global ICL. These results support approaching continual learning as a routing problem that chooses both a learning mechanism and a state cluster rather than committing to one representation globally.

Where You Inject Matters: The Trust Region Effect in LLM Agents

Chengquan Guo, Xiaodong Liu, Weiwei Yang, Mingqian Feng, Jialin Song, Qiyao Ma — Fri, 30 Oct 2026 00:00:00 +0000

90% · Evaluation 90%, Agents 83%

LLM agents are increasingly adopted in long-horizon tasks and interact with diverse external environments. Despite their growing capabilities, these agents remain vulnerable to real-world attacks such as indirect prompt injections, which manipulate agents through malicious instructions embedded in external environments. While existing risk assessment work has primarily focused on the injection content, the effect of where an injection appears has not been systematically characterized. Studying injection placement requires agents to operate over longer-horizon tasks and interact with diverse environmental surfaces. Thus, we build three agentic testbeds spanning tool use, coding, and daily workflow domains with diverse injection surfaces. Holding the injection content fixed, we find that agent susceptibility varies across injection placements. We refer to this phenomenon as the trust region effect: agents implicitly assign different levels of trust to different regions of their environments. For example, an agent powered by GPT-5.4 follows an unsafe injection instruction placed in a repository documentation file (i.e., a README or TODO file) 2.6 times more often on average than the same instruction placed in a code or data file. Building upon this observation, we develop learning-based attackers that generate injection content and select better injection placements. Our approach demonstrates that both where an injection is placed and what content it contains can affect the strength of indirect prompt injection attacks. To the best of our knowledge, this work provides the first systematic characterization for the effectiveness of different injection placements, spanning both coarse-grained environmental surfaces and fine-grained positions. The uneven trust that agents implicitly place across their environments provides important insights for future attack and defense design.

Interventional Training Reduces Causal Bypass in In-Context Introspection

Zilu Tang, Jongin Kim, Kerem Sahin, David Atkinson, Aaron Mueller, Derry Wijaya — Fri, 30 Oct 2026 00:00:00 +0000

90% · New model 90%, Evaluation 60%

Introspection, the ability of a system to access and reason about its internal signals, is crucial for language model (LM) self-consistency and scalable interpretability. Prior work argues that LMs possess introspective capabilities, and that training (SFT) could enhance such behavior. However, most setups leave open the possibility of "causal bypass," in which the self-report relies on signals that are correlated with the self-report's subject, while being causally unrelated to it. Although high-level concepts are difficult to intervene on, lower-level signals (such as a single neuron activation) allow precise interventions. Understanding such signals can serve as a minimal prerequisite before understanding more complex introspection. In this study, we train hundreds of introspectors across three low-level signal types and eight LMs. We find that straightforward SFT training leads to causal bypass, and propose interventional training, an alternative where we train the model to report on counterfactually perturbed values. This method enables more accurate introspection of endogenous $\textit{and}$ injected signals, particularly in out-of-distribution (OOD) data. The mechanism generalizes across token positions and scales with more data. Our results demonstrate that while causally grounded introspection is possible, it is not guaranteed -- emphasizing the need to train on mechanisms, rather than behaviors alone.

Distilling the Shape of Uncertainty: What On-Policy Distillation Learns and How to Predict It

Fri, 30 Oct 2026 00:00:00 +0000

90% · Evaluation 90%, New model 33%

On-policy distillation (OPD) is increasingly adopted in the post-training of large language models. However, a stronger teacher may fail to improve the student. Prior work attributes this to a lack of new knowledge, but what that knowledge is remains unclear. We show that OPD learns not only \textit{how to branch}, i.e., which candidate tokens to consider, but also \textit{where to branch}, i.e., where to keep reasoning paths open. The existing token-overlap metric captures only the former. We introduce the entropy shape, the sequence of next-token entropies, to characterize the latter, and propose the Entropy Shape Correlation (ESC) to measure student-teacher agreement in entropy shape. Tracking ESC throughout training and intervening on a teacher's entropy while preserving its token preferences, we empirically show that \textit{where to branch} causally drives OPD gains, independently of \textit{how to branch}. Finally, we formulate OPD outcome prediction as a teacher selection task and establish an evaluation protocol for it, under which pre-OPD ESC selects the best teacher for both students and outperforms existing predictors. Code is available at https://github.com/maybenotime/opd-esc.

Online Dense Reinforcement Learning with Planning-Grounding Agent for UI Navigation on Live IOS Apps

Peitong Duan, Khashayar Gatmiry, Jeffrey Nichols, Samy Bengio, Emmanuel Abbe — Fri, 30 Oct 2026 00:00:00 +0000

90% · New model 90%, Evaluation 66%, Agents 46%

While frontier vision-language models demonstrate strong capabilities in user interface (UI) navigation, their high inference costs prohibit everyday use. Conversely, smaller models trained via supervised fine-tuning (SFT) or offline reinforcement learning often memorize static trajectories, leading to brittle execution and an inability to recover from errors. To address these limitations, we propose a novel online reinforcement learning framework that trains UI navigation agents directly on live iOS applications. Our framework decouples the agent into two specialized components: a Planning agent and a UI Grounding model. We rely on an LLM-as-a-Judge to provide comprehensive reward signals, combining dense, step-level feedback to evaluate intermediate progress with a terminal reward for overall task completion. To effectively leverage this dual reward structure, we optimize the agent using Proximal Policy Optimization (PPO) with step-level advantage estimation. Furthermore, we introduce GiGPPO, a novel neighbor-conditioned value estimation method that improves critic stability by leveraging reward signals from visually similar states across concurrent trajectories. To support this online training paradigm at scale, we introduce an efficient infrastructure that orchestrates hundreds of concurrent virtual iOS devices. Evaluations on rigorous benchmarks, including UINavBench, iOSWorld, and AndroidWorld, demonstrate that our online RL trained agents substantially outperform both SFT and terminal-only RL baselines while accelerating training convergence. Notably, our 8B RL agents achieve nearly a sevenfold improvement over the SFT baseline in out-of-distribution generalization to Android environments, while outperforming Gemini-3.5-Flash and tying with Gemini-3.1-Pro on real iOS applications.

Neighborhood Elites: Quality-Diversity Search for Uncertainty Quantification in Question Answering

Mesut Ceylan, Arda Baris Basaran, Elgar Fleisch, Filipe Barata — Fri, 30 Oct 2026 00:00:00 +0000

90% · Evaluation 90%, Hallucination 82%

Large language models (LLMs) are increasingly capable, yet they produce wrong answers indistinguishable from correct ones. In question answering (QA), such an error propagates unnoticed into downstream tasks. Perturbation-based uncertainty quantification (UQ) addresses this by probing a model over a set of input rewrites and reading uncertainty from how much its predictions vary. Yet this set is usually built generically, by free paraphrasing, token-level perturbation, or adversarial search. Such rewrites leave two things uncontrolled that matter in QA: what kind of change is made to the context, and how explicitly the context supports the answer. When neither is controlled, the composition of the set is determined by the generator rather than by design, and the estimate inherits whatever that generator happens to produce. Quality-diversity (QD) search is built to cover a predefined exploration space, so we recast input perturbation as a QD problem and propose \emph{neighborhood elites (NE)}, a neighborhood constructor that spreads answer-preserving rewrites across two axes, the kind of edit and the explicitness of support. We compare NE against three other neighborhood constructors, holding the editor, QA model, uncertainty measure and search objective fixed, and against three scores read from the original context. Across five extractive-QA datasets and five instruction-tuned LLMs, NE improves failure detection against every other constructor in at least 24 of the 25 dataset-model pairs, reaching 0.726 AUROC and 0.211 AUPRC lift over the base rate. Additionally, NE detects failure better than semantic entropy computed over decoder samples, by 0.074 AUROC and 0.062 AUPRC lift, and this advantage widens as the neighborhood grows.

DolphinRouting: Cluster-Level Binary Routing for Disaggregated LLM Inference Serving

Yishuo Yuan, Shaoyu Yang, Chunrong Fang — Fri, 30 Oct 2026 00:00:00 +0000

90% · Efficiency 90%, Evaluation 39%

Prefill-decode (PD) disaggregation isolates the two phases of large language model serving, but request-level path selection must account for prefill work, cache reuse, and decode demand. We present DolphinRouting, which learns this choice from end-to-end SLO feedback under a given P/D resource allocation. Requests are clustered by raw prompt length, cache-adjusted input lengths on both worker pools, and output-length information. Offline policy search learns one binary action per cluster. Online feature extraction, cluster assignment, and table lookup select local or remote prefill. We evaluate DolphinRouting on a real PD-disaggregated cluster using three public benchmarks, a mixed workload, and real data. Across the mixed workload and real data, DolphinRouting with $k$-means reduces P90 TTFT by 5.7--7.9\% and P90 TPOT by 2.6--10.1\% relative to Dynamo.

★ TenetBench: A Clause-Anchored Russian Corpus for Evaluating Language Models against Professional Codes of Ethics

Thu, 29 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Codes of professional ethics guide and restrict specialists, who increasingly work with assistants built on large language models. Such an assistant can breach the user's code, or help the user breach it, without the user noticing. Evaluations reporting broad categories of harm cannot show how often this happens or which provision is breached. TenetBench pairs 3,433 requests conflicting with clauses of fifteen Russian-language professional codes with benign twins that differ only in the legitimacy of what is asked. An automated pipeline generates the tasks from the text of a code, so the corpus grows with every code added and a provision yields new tasks on every run. It also produced tasks and human-validated verdicts from two English codes of other jurisdictions. In human validation of the frozen pipeline, specialists rated 0.995 of the conflicting tasks plausible, and the model judge agreed with the adjudicated label at Gwet's AC1 of at least 0.894 on 1,860 elements. We evaluated fourteen model and reasoning-mode combinations on whether they produce the conflicting text, decline or warn. Eleven of the fourteen produce conflicting content on most such requests, mostly without a warning, and nearly three warnings in four accompany conflicting content. The models that conform more do so by declining while still serving the legitimate twin, and enabling reasoning halved the refusals of two models. We hope the dataset and pipeline will help evaluate how language models observe professional norms and let diagnostic datasets be built for particular professions and organisations.

★ EmbodiedLab: Benchmarking Embodied Agents That Act to Know

Hongcheng Gao, Zelin Zheng, Alex Zhou, Jay Zhu, Shijia Ge, Yazhe Wang — Thu, 29 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

Embodied agents are expected to accomplish real-world goals by deciding for themselves how and when to interact with their surroundings. For many such goals, changing the physical state of the world is the means rather than the end: the agent changes the state of a system through interaction, observes the feedback, and explores until it either achieves the goal itself (e.g., discovering physical laws) or learns what the goal requires; it acts to know. Manipulation benchmarks stop at acting to change, with the target state given; scientific discovery benchmarks evaluate knowing but replace physical experimentation with symbolic actions or numerical queries. We introduce EmbodiedLab, a benchmark in which a robot must act to know. Quantitative science experiments make this setting controllable: the parameters of the system are resampled for every instance, so the result cannot be known in advance, yet it can be scored exactly. Across 30 experiments in five disciplines on 14 apparatus platforms, the agent receives an objective and the apparatus, chooses its own procedure, and operates a physics-based simulated laboratory, reading the instruments through its cameras. Evaluation checks what the agent came to know, grounded in readings recorded during the episode, as well as the physical state the task requires; every experiment is verified to be solvable. We evaluate six frontier models in their developers' coding agents with native observation and control, with added perception and motion planning, and with manipulation skills distilled from the strongest agents. With native control, even the strongest agent succeeds less than half the time; for every agent, most failures arise in interacting with the apparatus, and manipulation skills help weaker agents more than perception and planning. The benchmark and its evaluation code are available at https://github.com/HexaFuture/EmbodiedLab.

How Far Can Coding Agents Go in Visual Generation?

Hengguang Zhou, Yuanhao Ban, Jiaqi Feng, Xiaohuan Pei, Justin Cui, Cho-Jui Hsieh — Thu, 29 Oct 2026 00:00:00 +0000

93% · Evaluation 93%

Modern image generation largely relies on specialized generative models that leave the process of image construction implicit in learned parameters. We explore an alternative: general-purpose coding agents that construct images through executable programs. Our study of zero-shot drawing on AgentDrawBench, a 200-prompt benchmark, reveals a distinctive capability profile. Agents excel at preserving structure and satisfying precise visual constraints, achieving competitive results in domains where compositional accuracy matters more than photorealistic appearance, while remaining limited in perceptual richness. Their visual expressiveness varies with the drawing interface, and a qualitative study on reconstruction shows how procedural skills extracted from human demonstrations support different artistic media. We then examine code as an external visual language of thought: a structured, executable representation that agents can construct and inspect. We test it in visual question answering task and find reconstructed visual code retains substantial information for question answering and, in some domains, supports higher accuracy than the original image. These abilities raise a further question: do agents draw on a shared internal representation of scene content beyond a particular drawing program? We explore an internal visual language-of-thought hypothesis, in which a shared, compositionally structured representation supports visual operations across drawing languages. Cross-harness drawings and images assembled from independent single-pixel queries exhibit related layouts and tones. Agents also maintain self-generated room arrangements across viewpoints after visible drawings are removed. These results demonstrate that the coding agent would maintain a consistent and coherent world inside its imagine even before the harness gets involved and the generation starts.

ProAS: A Progress-Aware Advantage Shaping Framework for Multi-Turn Agentic Reinforcement Learning

Bowen Ding, Mingda Wan, Yuhan Chen, Jiyao Yuan, Qi Zhu, Futing Wang — Thu, 29 Oct 2026 00:00:00 +0000

92% · New model 92%, Agents 91%

Reinforcement learning (RL) has emerged as the primary paradigm for optimizing Large Language Model (LLM) agents, yet efficient critic-free algorithms (e.g., GRPO) suffer from a fundamental uniform credit assignment dilemma that broadcasts trajectory-level advantages equally to all actions. Our analysis reveals that this uniformity inherently reinforces structurally unreasonable actions in preferred trajectories and ignores each action's downstream deviations in unpreferred ones. To resolve this without introducing critic overhead, we propose Progress-Aware Advantage Shaping (ProAS), a plug-and-play framework that leverages a dynamically evolving Agentic Reference Graph (ARG) to achieve fine-grained credit assignment via Masking Unreasonable Incentives (MUI) and Asymmetric Penalty Redistribution (APR). Extensive experiments across Qwen2.5-1.5B-Instruct and Qwen2.5-7B-Instruct demonstrate that ProAS universally elevates mainstream group-based baselines to new state-of-the-art performance, achieving peak success rates of 91.9% and 94.3% on ALFWorld, alongside 66.7% and 70.8% on WebShop, respectively. Furthermore, it accelerates training convergence by up to 2.44times while maintaining strong robustness against imperfect or distilled ARGs. Codes: https://github.com/LINs-lab/ProAS

VibeVideoBench: Evaluating Multimodal Agents on Code-Driven Video Generation

Muxi Diao, Wenbin Xu, Lele Yang, Xuyang Ma, Yuzhe Chen, Jiankang Chen — Thu, 29 Oct 2026 00:00:00 +0000

91% · Evaluation 91%

Diffusion video models produce striking clips, but they cannot reliably render an exact formula, a correct chart, a rule-consistent game or narration synchronized to on-screen events. A different route is emerging: multimodal agents that write code to render every frame and synthesize every sound. We introduce VibeVideoBench, a benchmark for evaluating agents on code-driven video generation. Its 239 self-contained tasks in English and Chinese span 12 categories, from data stories and music-driven pieces to classroom lessons and paper explainers; each asks an agent, running in its native harness inside a sandbox with no image or video generation model, to deliver a finished MP4 with sound within one hour. Evaluation is hierarchical and evidence-based: program checks of the delivery spec, about 13k task-specific rubric checks traced to the requirements they enforce, and five quality dimensions anchored per task, scored by three VLM judge panels. Across nine frontier agent systems and 2,151 scored rollouts, GPT-6 Astra and Claude Opus 5.5 rank first and second with both the judges and human annotators, and the strongest system satisfies every core requirement on 76.3% of tasks. Delivery is a second axis: counting missing videos as zero changes the ranking, and three quarters of the lost videos are final renders that started too late, ran too slowly or never started, a failure of how agents plan their hour. Agents build their own rendering pipelines and inspect their own frames, yet how much they look does not predict quality. The judges' ranking of the systems closely matches blind human preferences.

★ MedConstBench: Benchmarking How Constitutions Steer LLMs in Medical Ethics

Yibo Yang, Zerui Huang, Haomiao Yang, Zixia Wang — Wed, 28 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

As large language models advance rapidly, they are increasingly used to answer medical questions, and many such models are now deployed locally. Once a model is deployed, its behavior is governed mainly by the ethical policy written into its constitution. However, most existing work focuses on factual correctness rather than the trade-off between competing ethical values. Even benchmarks that do evaluate ethics treat it as a fixed property of the model rather than as a variable, so how a constitution reshapes medical-ethics behavior remains largely unexplored. We introduce MedConstBench, a medical-ethics benchmark that treats the deployment constitution as a controlled variable, is organized around Beauchamp's four principles, and is scored by human-calibrated multi-model peer review. It includes MedDilemma, our own dataset of clinical dilemmas, and supports a systematic analysis of how constitutions steer LLMs. Evaluating ten open-source LLMs under four constitutions, we find that the effect of a constitution is dimension-selective and model-dependent, and that no constitution is universally optimal. Motivated by these findings, we propose a do-no-harm adaptive selector that chooses a constitution per model, matching the per-model oracle while guaranteeing a worst-case floor and generalizing to unseen models at low cost.

★ Beyond Update Counts: A Control Map for Evaluating Test-Time Model Updates

Zehao Zhao, Shangjian Yin, Rui Liu, Chongyang Bai, Mark Ibrahim, Kavosh Asadi — Wed, 28 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Test-time updating lets a language model change persistent state after training---such as model weights, adapter parameters, or a memory module---before or while answering queries. Existing evaluations often compare systems using update count or frequency and performance measured immediately afterward. These summaries omit what is updated, where, when, and in what order. The same total can therefore hide an important difference: one system may answer after one update while another answers after two. We introduce the Control Map, an evaluation framework that records the committed updates preceding each scored answer and separately evaluates behavioral changes, immediate learning (acquisition), later recall (retention), preservation of unrelated behavior (locality), and performance in new settings. We prove when counts alone determine final state and show, under stated assumptions, how differences between pre-answer histories limit differences in average scores. Using ControlMap, our experiments reveal three insights. First, update scheduling matters: at a 760M-parameter checkpoint, 24 schedules with the same update count yield 19.7%--34.2% accuracy. Second, learning an edit does not guarantee that it lasts or remains contained: all six selected CounterFact LoRA settings learn the new information, while ROME retains edits at all 13 operating points but fails locality at each. Third, no editor performs best consistently across models, benchmarks, and tests. Evaluations should therefore report update histories and assess acquisition, retention, and locality separately.

★ How Stable Are LLM-Generated Proofs Under Program Refactoring? A Stress Test for Verifiable Code Generation

Aravind Mandiga, Guoming Li — Wed, 28 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Benchmarks for verifiable code generation usually test a proof against one fixed implementation. We ask whether the same proof remains valid after a program refactoring that preserves behavior. We introduce RefactorProof, which derives 458 Lean-certified variants from the 185 VERINA tasks that compile in our pinned environment while holding the specification and evaluated proof fixed byte-for-byte. VERINA supplies reference proofs for 46 tasks, all from its basic split; the primary model-versus-reference comparisons further match on 21-25 short, non-recursive tasks. Structural refactorings expose substantial brittleness across proof sources: task-normalized structural PSR is 4.3% for the supplied references and 13.3-26.8% for the specialized provers. On the matched tasks, specialized prover proofs nevertheless have higher overall unchanged-proof survival than the supplied reference artifacts: +15.6 points for DeepSeek-Prover-V2-7B, +18.8 for Goedel-Prover-V2-32B, and +20.3 for Goedel-Prover-V2-8B; all three 95% CIs exclude zero. The ordering persists after excluding universally survived T1 variants, under family-balanced aggregation, and under alternative deployable valid-proof selection rules. However, unchanged-proof failure is not the same as high repair cost: predefined one-identifier edits repair 51/54 eligible reference structural failures and 58/85 eligible prover failures. RefactorProof therefore measures unchanged-proof stability under program refactoring, not the cost of maintaining a proof after it breaks.

★ kw-bench: Breaking Frontier Harnesses on Real-World Tasks

George Juraj Salapa — Wed, 28 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 90%

Each new generation of frontier model pushes the edge of what is possible. Inside a harness -- a loop, a shell, a filesystem, the ability to write and run its own code -- a model can take on increasingly difficult work. A growing body of work treats the coding-agent harness as infrastructure for knowledge work rather than a coding tool, and a new wave of benchmarks tests it there. Most of them are generative: build a financial model, a memo, a spreadsheet from a brief, judged by a rubric or by another model. Few test the expensive part of knowledge work, which is investigative. In audit, tax, due diligence, regulatory reporting, claims or any dispute over numbers, the answer already exists -- hidden in incomplete and inconsistent evidence spread across systems -- and has to be recovered and defended to the cent. We introduce kw-bench, a benchmark built from real engagements specialists described to us. Its first task family is a five-year VAT reconstruction across three generations of SAP -- legacy ECC, a partial S/4HANA migration and a consultant's redesign -- delivered as raw system extracts: page headers, local number formats, partial link tables, no documentation of how anything connects. Grading is deterministic, with no LLM judge. We ran four harnesses -- Claude Code, Codex, OpenCode and micro-cc -- on the same model, DeepSeek v4.1 Flash, for two hours. The two independent, model-agnostic harnesses delivered (OpenCode 0.542, micro-cc 0.528); the two lab harnesses, away from their home model, delivered nothing (0.00 each). What defeated them was inertia: keeping many weak signals alive across a long session. On a frontier model the task collapses. micro-cc and Claude Code on Claude Opus 5.5 scored 0.9986 and 1.000, in 22 and 25 minutes. A weeks-long engagement now costs minutes and a few dollars; below the frontier, the harness decides how far a model gets.

★ When to Intervene? An Empirical Study of Errors and Self-Recovery in LLM Agents

Chiming Duan, Tong Jia, Yingxi Chen, Minghua He, Lingzhe Zhang, Ying Li — Wed, 28 Oct 2026 00:00:00 +0000

98% · Agents 98%, Evaluation 95%

LLM agents interact with Web services to perform tasks on users' behalf. Errors during these interactions can disrupt execution and prevent task completion, motivating external intervention. However, agents may self-recover through continued execution, raising the question of when to intervene and when to wait for self-recovery. We present an empirical study of errors and self-recovery in LLM agents. We construct ReInBench, a dataset containing 1,259 agent traces and 1,403 error events across computer-use and Web tasks. Our study examines how often and how quickly agents self-recover, how self-recovery relates to subsequent execution and task completion, and whether self-recovery can be predicted when an error occurs. We find that 725 error events self-recover, often within a few steps, with self-recovery performance differing across Execution LLMs. Self-recovery can also help agents recover from subsequent errors of the same type and support overall task completion. Intervention should therefore prioritize errors unlikely to self-recover, while allowing agents to continue when self-recovery is likely. Building on these empirical findings, we propose ReCoF, a method that predicts whether an error will self-recover to inform whether additional intervention is needed. ReCoF achieves an average relative improvement of 9.02% in Macro-F1 over the strongest baseline on ReInBench. Our code and data can be found at https://method-notes.github.io/ReCoF/.

★ CogArena: Benchmarking Multimodal Agents on Interactive Behavioral Experiments

Wed, 28 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 92%

Claims about the cognitive capabilities of large language models are typically tested by translating behavioral paradigms into text prompts, bypassing the visual interfaces and time pressure through which human participants actually engage with experiments. At the same time, AI agents can now complete online behavioral tasks and produce human-like data, which raises concerns that online datasets in the social and behavioral sciences are increasingly open to contamination. We introduce CogArena, a benchmark of ten experimental paradigms from the social and behavioral sciences. Agents complete the experiments through a standard web browser, working with the same visual stimuli and response interfaces that online participants use. A three-level scoring procedure evaluates whether the agent completed the task, how accurately it performed, and whether its data show the behavioral signatures each paradigm reliably produces in people. Across four multimodal models, each run ten times per task, we find that finishing an experiment and reproducing its signatures are different things. Many sessions did not finish, and those that did reproduced some signatures reliably and others not at all. CogArena provides the tasks, scoring code, and a public leaderboard for new submissions.

★ Benchmarking Task Takeover Across Agent Handoffs

Wenyuan Li, Guang Li, Dongyuan Li, Keisuke Maeda, Takahiro Ogawa, Miki Haseyama — Wed, 28 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 87%

As agents collaborate on longer tasks, responsibility may pass between models before the work is complete. The receiving model must build on prior decisions, observations, and actions while addressing the remaining requirements. Evaluating this transition requires connecting the information preserved at handoff to how the receiver continues the work, a relationship that task-success and memory-retention scores alone do not capture. We introduce the Handoff State Continuity Benchmark (HSC-Bench), comprising 1,039 tasks that evaluate two complementary aspects of task takeover: recovering relevant repository evidence and continuing from established conversational, tool, and file-work state. Comparisons across six receiving models show that structured handoff state improves repository evidence recovery for every model and continuation performance for most models. The strongest receivers differ between the two tracks, showing that evidence-recovery rankings alone do not characterize performance when continuing from inherited state. By combining these comparisons with checks of subsequent responses and actions, HSC-Bench evaluates both access to prior evidence and adherence to established commitments and completed work. This provides a basis for selecting receiving models and assessing handoff representations according to the work that must be carried forward.

★ ETC-Bench: A Multilingual Benchmark for Evaluation-Side Transcript Canonicalization

Zhiheng Qian, Junhao Du, Jing Peng, Hanqi Li, Kunyang Peng, Feng Xue — Wed, 28 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Written transcripts of the same speech can differ in ways that are not uniquely determined by the speech signal. We formulate evaluation-side transcript canonicalization (ETC), which maps multiple surface forms to a single canonical form when their differences are not determined by speech. To our knowledge, ETC-Bench is the first and most comprehensive multilingual benchmark for evaluation-side transcript canonicalization, covering 181 phenomenon types across 12 languages. Each instance is a minimal pair that shares a sentence context and differs only in a target span exhibiting one phenomenon. To evaluate both canonicalization success and semantic preservation, we propose Semantic-preserving Exact-match Normalization Accuracy (SENA), which combines exact agreement between independently canonicalized pair members with a pair-relative BERTScore constraint against information loss. We further introduce the Taxonomy-guided Agent Canonicalization Skill (TACS), which provides LLM agents with phenomenon descriptions and explicit canonicalization contracts. Experiments with five baseline configurations and three TACS-based agents show that the agents outperform the baseline systems on shared languages. ETC-Bench provides a systematic testbed for evaluation-side transcript canonicalization across languages.

★ InfraDev: Can Coding Agents Develop AI Infrastructure, and How Can We Tell?

Yangshen Deng, Jiaqi Ruan, Yeqi Huang, Nick Lau, Guangming Sheng, Luo Mai — Wed, 28 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 37%

Coding agents are beginning to develop the production-grade AI infrastructure that trains and serves large models, yet existing benchmarks cover only single kernels or single-GPU tasks, and grade with tests and prompts that misjudge both correct and incorrect patches. We introduce InfraDev, a benchmark of 36 tasks built from real-world pull requests to Megatron-LM, vLLM, and verl, covering infrastructure development for training, inference, and reinforcement learning. Each task asks an agent to reimplement a real change that spans many files and must run correctly across many GPUs. To grade such changes, we jointly synthesize hidden probes, which enforce the requirements the original tests leave unchecked, and a prompt that states every interface the probes call without revealing the solution. We also design a shared execution environment that reuses dependencies and artifacts across tasks, cutting storage by 52-57% and build time by 13-14% on Megatron-LM and vLLM. Across six frontier models, the best resolves only 38.9% of the tasks; the probes reject 32% of the submissions that pass every official test, and with the original pull request descriptions as prompts, no model resolves a single task. Failure analysis shows that structural and state errors dominate and that agents' own tests rarely check memory, time, or traffic. Turning these findings into an agent skill raises the resolved rate of a frontier model on the Megatron-LM tasks by 10.5%.

★ AutoVend: Benchmarking LLM Agents on Long-Horizon Autonomous Vending Operations

Qiming Shi, Yikun Li, Beidi Luan, Rui Sun, Yan Ma, Jing Li — Wed, 28 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 90%

Large language model (LLM) agents are extending agent-mediated commerce on the Web from shopping for consumers to operating businesses for sellers. In online-to-offline (O2O) commerce, decisions are issued through Web-based platforms but realized through costly physical execution. Evaluating such agents requires going beyond individual transactions to ask whether decisions made online turn into sustained economic returns. We introduce AutoVend, a benchmark in which one LLM agent manages 20 vending machines across five site types through a Web-based operating interface for 180 simulated days. Its data-grounded simulator combines a shared warehouse, procurement lead times, paid replenishment, inventory expiration, and recurring operating costs. We evaluate five LLM agents against a rule-based inventory-control baseline and diagnose their site coordination, execution, and performance over time. Four of five evaluated LLM agents fall short of this baseline, and returns per dollar of inference cost rank models differently from returns alone. Operational analyses show that gains in individual metrics do not consistently yield stronger economic growth, as inventory and execution costs erode these gains. Trajectory analysis further reveals difficulties in grounding feasible actions, managing inventory commitments, and converting feedback into effective policy revisions.

★ OpenGray: a benchmark for evaluating the decisions of radiotherapy planning agents

Humza Nusrat, Kundan Thind — Wed, 28 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Background and purpose: Radiotherapy planning agents based on large language models (LLMs) decide when to submit a plan or request review but are evaluated mainly by dose. This study developed OpenGray, a benchmark of planning tasks with known correct decisions. Materials and methods: OpenGray adds tasks with known answers to the OpenKBP-Opt head-and-neck dataset, including two contradictions with dose goals proven incompatible and a feasible control with a stored plan meeting every hard goal. Four systems, from a rule-based planning script to a full LLM agent, were evaluated on 30 patients, the LLM-based ones with a commercial (Terra) and a local (Ministral 3 8B) LLM. Decisions, achieved dose, problem recognition, and technical failures were recorded and also read from logs of 39 stereotactic radiosurgery runs of SAGE, a clinical LLM agent. Results: With Terra, the script preceded by an LLM review and the full agent had decision accuracies of 95.4% and 92.1%, respectively, yet escalated only 50% (95% confidence interval, 25-75%) and 25% (10-40%) of overlap contradictions, which required arithmetic on the anatomy. All four systems submitted every feasible control, but 24% of these plans missed a hard goal that the stored plan had met. In four SAGE runs, one goal failed at every attempt; the agent identified it in every reply but had no action for stopping. Conclusions: Favorable aggregate scores concealed specific decision failures. Planning agents should be evaluated on their decisions as well as their dose and given an explicit action for stopping and requesting review.

★ LiveSearch: Questions as Executable Verifier Code for a Live Benchmark of Professional-Source Financial Search Agents

Sinuo Wang, Wentao Zhao, Jiacheng Lu, Rui Sun, Beidi Luan, Qiming Shi — Wed, 28 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 97%, Hallucination 37%

In finance, the answer that matters is usually the current one: prices, rates, and disclosures hold only as of a given day, so retrieving up-to-date facts is a core capability of a financial agent. Yet most financial benchmarks overlook timeliness. They fix each answer at annotation time, so the gold goes stale as markets move and, once leaked into training data, rewards recall rather than search and analysis. We introduce LiveSearch, a live benchmark that stores each question as executable verifier code: a typed program of calls to a professional financial data service, followed by a declarative extractor. The question is frozen, while its gold answer is determined by executing the program against professional financial data at evaluation time, and a daily job re-executes it with no human or language model in the loop. A memorized answer is thus likely outdated, and a model must reach the current one through tools and analysis. Construction is equally automatic: a language model only verbalizes programs enumerated from real tool responses and never chooses an answer. Under general web search, six contemporary tool-calling models reach 70.9% to 76.7% answer accuracy. Over three consecutive days, 70% of served questions change their gold value, and even under tolerance-aware matching, a literally reused earlier snapshot fails on a majority of questions within a single day. Per-question tracking uncovers dynamic shifts in correctness masked by aggregate accuracy.

★ TCM-DiagCommit: A TCM-SD-Derived Benchmark for Evidence-Conditioned Syndrome Selection and Commitment

Erwei Hou, Yue Hu — Wed, 28 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 47%

Traditional Chinese Medicine (TCM) syndrome differentiation is usually evaluated as static final-label selection over complete clinical records. This setting hides three diagnostic behaviors: whether a model distinguishes semantic-neighbor syndrome labels from easy random distractors, whether it updates predictions as evidence fields are revealed, and whether it should commit under uncertainty. We introduce TCM-DiagCommit, a benchmark derived from the public TCM-SD/ZY-BERT resource while retaining its source syndrome labels; it converts 1,834 source cases into 23,842 task rows over 148 syndrome labels. The benchmark combines random, profile-neighbor, and full-vocabulary candidate ladders with staged t1/t2/t3 evidence. It also includes fixed-label tongue/pulse perturbations and insufficient-evidence variants. A benchmark-defined action contract scores commit, acquire, defer, and abstain. Across five zero-shot DirectLLM calibrations, random10 top-1 accuracy is 63.2% to 66.4%, but hard20 falls to 24.7% to 30.3% and full148 to 14.8% to 22.3%; on hard20, WT(all) wrong-terminal commits remain 62.2% to 74.3%. These results show that final-answer accuracy alone is insufficient for diagnostic benchmark evaluation: answer selection, evidence use, and benchmark commitment scoring should be measured separately.

★ M4Bench: Evaluating Procedural Specification for Clinical EHR Derivation Agents

Wed, 28 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Clinical research over electronic health records depends on derived variables whose definitions sit outside the schema, in item codes, time windows, missingness rules, and output-grain choices that change cohorts and downstream conclusions. We introduce M4Bench, 28 clinical derivation tasks across 15 task families on MIMIC-IV and eICU, to test whether AI coding agents can reproduce such derivations and how clinician-reviewed agent skills change their results. Its controls decompose the value of skills into specification completeness, delivery format, and schema familiarity. In a Docker-isolated Codex campaign (GPT 5.4 Mini and GPT 5.5), skills raise reward by 0.136 over the no-skill baseline, with 26 of 28 tasks improving. Specification completeness accounts for most of this gain: on an eight-task control subset, an operational spec alone raises reward by 0.162, and skills add only 0.024 on top, indistinguishable from zero. On a ten-task subset, decoy skills score like no skill and raw reference SQL nearly matches targeted skills, arguing against generic context or the skill format as the explanation. Schema obfuscation and restructuring preserve 86% and 70% of the skill gain (GPT 5.4 Mini, 26 MIMIC-IV tasks), a descriptive robustness check rather than a bound on memorization. A descriptive Claude Code replication on all 28 tasks (Opus 4.7 and Sonnet 4.6, 560 runs) is directionally consistent (+0.120, 21 of 28 tasks improving). These gains are partial agreement with the reference, not completed tasks: only 25 of 280 with-skill and 13 of 280 no-skill Codex runs pass every strict check, and skills barely change which rows agents return (key F1 +0.003). Skills are released as a versioned, clinician-reviewable markdown delivery format, with a family-level reference-validity audit. M4Bench releases tasks, skills, evaluation code, and sanitized per-run evidence as an open-source benchmark.

★ Gym-Anything-RT: Scaling Real-Time Environments for Vision-Language Agents

Pranjal Aggarwal, Sean Welleck — Wed, 28 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

AI agents have made substantial progress in domains where they have a non-trivial amount of time to deliberate between actions, such as mathematics, software engineering, and performing tasks on desktop applications. It is less clear how such models fare in settings that demand real-time interaction, such as playing a video game or operating a physical device. In these settings, the environment state may evolve as the agent is reasoning, or even as the underlying model is performing a forward pass. To gain insight into the real-time interaction capabilities of vision-language agents, we introduce Gym-Anything-RT. Our key idea is to construct visual minigames that demand a portfolio of capabilities required for interactive real-time settings. Gym-Anything-RT uses an agentic pipeline to automatically generate such environments based on a taxonomy of six capabilities required for real-time interaction, including the ability to perceive the environment and act under an evolving environment state. A key feature of Gym-Anything-RT is that it allows for controlled comparison of real-time interaction capabilities: we can run the same minigame in real time, or pause or slow down the environment to relax the requirement for real-time actions. We use Gym-Anything-RT to construct 100 real-time minigames and evaluate AI agents on them. We find large capability gaps in real-time interaction. For example, GPT-6 Astra (xHigh) achieves a pass rate of only 9.0% on the real-time minigames, which rises to 75.0% when the environment is paused. We also identify intriguing behaviors by the most powerful frontier agents: for instance, when given the ability to write code, these agents write code that bypasses the need to perceive and act in real time. We publicly release Gym-Anything-RT and the generated minigames to spur further research on AI agents in real-time interactive settings.

★ Rare, Relevant, and Still Wrong: How Memory Fails LLM Group Decisions

Anouar Imel, Imad Charradi, Abderrahman Sghir, Khalil El Khammar — Wed, 28 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

When an LLM group's discussion is compressed into a short memory for a final decision maker, what should the memory keep? Two natural heuristics keep the rarest claims or the claims relevant to the decision criteria. On hidden-profile tasks both look attractive, because how widely a fact is held is confounded with whether it helps. We introduce CrossProfile, a diagnostic benchmark of 60 templates with automatically verified answer keys, in which the ownership of supporting, decoy, and irrelevant facts is crossed while their text is held fixed (480 tasks). Across Qwen2.5-7B, Llama-3.1-8B, and two DeepSeek models, both heuristics fail in predictable places. Selecting by relevance readmits decoys, because decoys are on topic: even a perfect relevance filter loses 23 to 31 points to rarest-first selection in the classic hidden-profile configuration. Rarest-first selection is no better than random selection on average and wins only where the rare facts happen to be the useful ones. In both cases accuracy follows the balance of evidence the memory keeps, not the policy that kept it. The decision model adds a third failure, a form of social proof: it gives extra weight to facts attributed to many agents. Removing the attribution tags, or reassigning them at random with line order fixed, raises accuracy in the classic configuration by 13 to 24 points. In natural discussions the evidence is lost before any memory acts: the complete transcript yields lower accuracy than no notes at all.

★ Scaling Laws for Tool-Use Agents via Controllable Environment Synthesis

Wed, 28 Oct 2026 00:00:00 +0000

98% · Agents 98%, New model 77%, Evaluation 70%

We explore how to build performant large language model (LLM) tool-use agents that complete complex, multi-step tasks. Specifically, we study how data composition in synthetically generated tool-use environments affects the training of such agents. To do so, we propose the SynthAPI pipeline, which explicitly controls the diversity of generated environments along the axes of tools, tasks, and the databases on which the tools operate. By systematically varying SynthAPI's generation controls in 458,752 tasks across 849 trained models of four different sizes, we derive a scaling law that predicts downstream tool-use agent generalization based on the scale and composition of the synthesized environments. These predictable relationships reveal several lessons for scaling tool-use agent data. As data budgets grow, scaling tool and database diversity becomes more effective than increasing the number of tasks, and increasing data complexity yields larger gains. Scaling according to these findings yields agents that outperform models trained on data from prior work in large-scale tool-use agent and synthetic environment generation across four evaluation benchmarks.

★ CodeCap-Bench: Can Code Serve as a Good Video Caption?

Jiafu Tang, Yifan Xu, Eric Tang, Junjie Zhao, Shipeng Luo, Te Sun — Wed, 28 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 30%

Video captioning aims to preserve visual information through compact descriptions, predominantly using natural language. However, precisely expressing fine-grained spatial structures and temporal dynamics in natural language often leads to lengthy and potentially ambiguous descriptions. In this work, we explore Code as Video Caption, a paradigm where executable programs serve as a complementary representation of visual content through graphical primitives, layout rules, and functions of time. To examine whether code captions effectively preserve video information, we compare them with natural language captions on downstream question answering and video reconstruction tasks. Our results show that code captions achieve superior reasoning accuracy and reconstruction fidelity. Despite these advantages, realizing the potential of code captioning depends on whether current multimodal large language models (MLLMs) can generate faithful code captions. To systematically assess this capability, we introduce CodeCap-Bench, which comprises 323 videos spanning five categories of structured 2D content, along with fine-grained annotations and a multi-level evaluation framework. Extensive experiments reveal that while leading proprietary MLLMs demonstrate promising code captioning capabilities, they still struggle to faithfully capture complex temporal dynamics, with open-source models lagging substantially behind.

★ Instruction Capacity: Behavior and Degradation of Language Models under Controlled Load

Hennadii Kutomanov — Wed, 28 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

The quality with which language models follow instructions falls as the load grows, yet there is still no quantity that expresses the limit of a model in units of load with an uncertainty. We propose an instrument in which the load consists of rules that replace single-token words in a text without meaning: such an instruction is atomic, the rules are independent by the construction of the task, and the outcome of every application of a rule is determined by a computable reference answer. Every occurrence of a word to be replaced becomes a Bernoulli trial with a known position in the answer and a known rule number. The statistical model first measures the shape of the dependence of the probability of a right replacement on the text length, the word position, the replacement density, the number of rules and the rule number, then describes it by smooth laws and restores a reference level that does not depend on the choice of the coding. The model yields the instruction capacity - the number of rules held at a given share of right replacements - with an uncertainty. A measurement of GPT-4o mini (2958 requests, 642,430 trials) gave regular laws for every load factor and for the collapse of the answer; the instruction capacity was about 37-38 rules at a share of right replacements of 0.95, and about 18-19 rules at the full replacement density. The code of the instrument, the measurement data and the passport are open, and every number of the passport is recomputed from the published points. The passport of a model serves as the calibration of a future meter of the instruction load of prompts.

★ CASEL: Consequence-Aware Selective Execution for Reliable Long-Horizon Tool-Using Agents

Yanyan Zhang, Fei Shen, Ting Yan, Hao Zhang — Wed, 28 Oct 2026 00:00:00 +0000

97% · Agents 97%, Evaluation 66%

Tool-using language agents increasingly modify external systems, yet most runtime safeguards allocate scrutiny by generic uncertainty scores or coarse mutation heuristics. We introduce CASEL (Consequence-Aware Selective Execution Layer), a model-agnostic runtime allocation policy that combines calibrated failure probability, state-dependent consequence, reversibility, ambiguity, and evidence freshness to choose, for each proposed tool call, among direct execution, verification, sandboxing, clarification, approval, and abstention by minimizing a per-mode cost. We formalize direct execution as selective prediction and state and test the validity conditions of a finite-sample threshold and a trajectory risk ledger. In a controlled simulation over 20 seeds with 120,000 test actions each, in which every method has the same intervention menu and the same compute cost, CASEL reduces harmful-side-effect trajectories from 14.75 percent for an uncertainty-ranked gate to 7.02 percent and mean normalized harm from 0.184 to 0.123. A gate ranked by the product of failure probability and consequence reaches similar harm (5.82 percent, 0.120) but sends 45 percent more actions to human approval, and removing reversibility-aware routing raises the rate to 15.53 percent. A trajectory-level risk budget, in contrast, increases harm at equal cost (11.82 percent). Intervention effectiveness is assumed, not measured: the results rank allocation rules under the simulator's assumptions and do not establish deployment effectiveness.

★ PsyJev: Psychological Analysis via System-1 Judgment

Fangyuan Liu, Sirui Zhao, Xiaohao Wang, Guo Cheng, Peng Zhang, Enhong Chen — Wed, 28 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, New model 94%

Self-disclosures on the Web support diverse psychological analyses, from recognizing expressed experiences to identifying symptom-relevant information. We study a shared, question-conditioned decision interface for this heterogeneous problem in content analysis. We organize PsyJudgeBench, comprising 13 annotated datasets and 82 questions across six psychological areas, while preserving source-specific criteria and masks for unobserved labels. We introduce PsyJev, a System-1 model that directly produces distributions over answers specified in each request. Joint supervised training establishes the shared judgment function, followed by Judgment-Anchored RLCD (JA-RLCD), which adapts logit-space reward optimization to heterogeneous psychological judgments through balanced supervision, selective retention of correct SFT predictions, and question-group reward optimization. The Qwen3.5-based 2B, 4B, and 9B models achieve 64.33, 66.21, and 67.56 average test F1, respectively. The 9B model exceeds the strongest external comparison by 2.54 points. Controlled 2B comparisons examine the training stages and anchored refinement. Further analyses examine question-conditioned judgment, inference cost, and probability quality, complemented by qualitative comparisons across models.

★ DocSkill: Skill-Integrated Agentic Reasoning for Multilingual Multi-Hop Document Question Answering

Jing Huang, Zhiya Tan, Changtao Miao, Shutao Gong, Yang Wang, Joey Tianyi Zhou — Wed, 28 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, New model 75%, Agents 53%

Multilingual multi-hop document question answering (MMHop-DocQA) requires locating and integrating evidence across complex layouts, document pages, and diverse languages. Existing approaches include single-pass processing and adaptive tool use, but learning to coordinate evidence acquisition, cross-page integration, and answer verification remains challenging. To address this challenge, we present DocSkill, an agentic framework with a learned skill-orchestration policy for adaptive evidence acquisition and reasoning. The policy selects, revisits, or skips skills according to the evolving evidence state. To construct supervision for skill orchestration, bounded Monte Carlo Tree Search (MCTS) explores alternative skill trajectories offline. These trajectories provide demonstrations for supervised fine-tuning (SFT) and preference pairs for Direct Preference Optimization (DPO). We further introduce MMHop-DocBench, a multilingual multi-hop document QA dataset with training and test splits, containing 60K QA pairs across 10 languages and eight document domains. Its HopGen pipeline builds hierarchical page-element graphs to derive traceable cross-page evidence paths and generate grounded QA pairs with page- and element-level evidence annotations. Experiments show that DocSkill with Qwen3-VL-8B achieves 41.0 accuracy and 49.2 F1 on MMHop-DocBench, exceeding the strongest evaluated baseline by 5.5 and 6.3 points, respectively, and achieving competitive performance on MMLongBench-Doc and DocBench.

★ Beyond the Evidence Coherence Trap: Agent-Driven Evidence Selection for Temporally Disordered Fake News Detection

Chi Wang, Min Gao, Yifei Zhang, Bincheng Gu, Zongwei Wang, Wentao Li — Wed, 28 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, New model 54%

Fake news undermines public trust and the integrity of online information ecosystems. Temporally disordered fake news is especially elusive because it rearranges factual events into a false chronology. Individual events may remain factually valid even when their temporal relations are misrepresented. When similar events involving the same entity recur, retrieval may return semantically plausible evidence from the wrong period. Consequently, independent event verification may rely on locally coherent but temporally mismatched evidence and incorrectly validate the article. We call this the \textit{evidence coherence trap}. Because existing benchmarks rarely isolate this setting, we construct ChronoFake, a source-disjoint benchmark of 30,706 articles derived from temporally anchored real events. It distinguishes temporal manipulation from valid narration, including flashbacks. To address this, we propose DelSA, a \underline{D}ocument S\underline{el}ection \underline{S}elf-Updating \underline{A}gent that iteratively refines evidence selection during training and applies a frozen rulebook for one-pass evidence selection at inference. DelSA uses an LLM to characterize candidate evidence, while a verifier-trained reranker balances local relevance against timeline consistency. Finally, the classifier compares the article-implied chronology extracted by the LLM with the evidence-derived chronology to reveal global temporal conflicts. DelSA outperforms existing methods on ChronoFake, while remaining competitive on PolitiFact and GossipCop.

★ RubricAtlas: Mining Type-Level Rubrics from Expert Reports to Evaluate Deep Research Agents

Delin Wang, Sinuo Wang, Rui Sun, Beidi Luan, Qiming Shi, Jing Li — Wed, 28 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 96%

Deep research agents conduct multi-step Web searches and synthesise retrieved information into analytical reports. Evaluating these outputs requires checking which analyses they deliver, beyond the presence of relevant facts. Query-specific binary rubrics make individual requirements explicit, but task-level checklists may leave the overall analytical structure underexamined. We introduce RubricAtlas, which derives type-level analytical requirements from expert reports on different objects within the same research type and organises detailed checks around recurring analytical questions. The requirements give related tasks a shared standard for what to analyse, without prescribing their answers. Our framework links binary checkpoints to report passages and evaluates completeness separately from argument validity. Built from 3,079 Chinese financial research reports, RubricAtlas covers 99 research types. Reference-report selection experiments show that individual reports cover only about half of their type's analytical questions, and that selecting requirements from different reports changes evaluation scores and agent rankings. Across ten agents, two judge pipelines achieve checkpoint agreement of $\kappa$ = 0.71 and agent-ranking agreement of Kendall $\tau$ = 0.87. Controlled deletions reduce scores on the affected analyses, while padding produces no detectable increase. These findings support type-level rubrics for comparing the analytical completeness of Web deep research outputs, with individual decisions traceable to report evidence.

★ Proactive Memory for Event-Driven LLM Agents

Jinyang Li, Zichuan Fu, Xian Wu, Wenlin Zhang, Yimin Deng, Kaifeng Guo — Wed, 28 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Proactive LLM agents anticipate service opportunities from ongoing observations and accumulated history. Existing approaches emphasize recent context or retrieval of related memories, providing limited support for validating recurring behavioral regularities and assessing their applicability to the current event. We propose Proactive Memory (ProactMem), which organizes history into recent events, temporal trajectories, and behavioral regularities. Trajectory facts are checked against source events, and regularities are assessed through historical replay before being matched to current conditions. We also construct ProactiveStream, a longitudinal benchmark covering three application scopes, where agents process isolated event streams chronologically under episode-aware opportunity labels. On ProactiveStream, ProactMem achieves an overall effectiveness (F1) of 52.6%, a 21.4% relative improvement over the strongest of five external baselines. Code is available at: https://anonymous.4open.science/r/ProactMem-5351/.

★ TeamGym: A Gym for Team Decision-Making by AI Agents with Distributed Information

Wed, 28 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, New model 36%

Many real-world decisions require teams to integrate information distributed across participants. Yet AI agents are still primarily trained and benchmarked in isolation. We introduce TeamGym, an environment for evaluating and training language model-based agents on team decision-making problems constructed from real-world legal, finance, medical, and software engineering corpora. TeamGym introduces a method to identify a minimal set of facts, each element of which is necessary and all of which are collectively sufficient for a correct final decision, which we call pivotal information. This enables TeamGym to both control the distribution of pivotal information and track how well agents identify and integrate pivotal information when making team decisions. Examining 8 language models on 12 source corpora, we find that current AI agents often fail to effectively share information: Distributing pivotal information across participants lowers success rate by 22.9 percentage points on average relative to giving the same information to a single model (a communication gap). We also document that participants do not effectively retrieve decision-relevant information on their own: A team's success rate is lowered by a further 19.9 points when participants must identify and extract pivotal information from source material rather than receiving it directly (a retrieval gap). To help remedy these deficits, we use TeamGym to train policies through reinforcement learning on team decision loss. Across two policy-training experiments, we find that training improves final-decision quality and induces emergent coordination in participation, speaker selection, and communication, even though the reward evaluates only the final decision.

★ Let the Gate Decide: Per-Lookup Precision for Quantized Conditional Memory in LLMs

Tianqi Chen, Xunchao Li, Linchuan Xie, Guanghua Yu, Jianchen Zhu, Peisong Wang — Wed, 28 Oct 2026 00:00:00 +0000

96% · Efficiency 96%

Conditional memory layers, such as DeepSeek's Engram, attach large hashed n-gram tables to a language model, and removing them from DeepSeek-V4.1-Flash costs 36.8 PopQA points. Low-bit quantization loses long-tail facts first, and the standard remedy, giving more bits to frequently accessed rows, backfires: refining lookups by access frequency is worse than refining them at random. We trace this to the gate that scales each retrieved row. A lookup's quantization error is weighted by its gate, which is small for frequent n-grams, large for the rare n-grams of long-tail entities, and computable from a 1-bit code. GateRQ therefore stores each row as a ladder of 1-bit residual codes, keeps level 1 in GPU memory to compute the gate, and reads deeper levels only for the lookups the gate ranks highest. Without retraining, GateRQ cuts GPU table memory by 7.8x on DeepSeek-V4.1-Flash, which raises vLLM throughput by 11%-31% over FP8 tables, while losing about 1 PopQA point at 1.50 bits read per lookup and 0.2 points at 2.10 bits across two models.

★ From Generation to Scoring: Scaling LLM-Based Semantic Labeling at LinkedIn

Wed, 28 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, New model 66%, Evaluation 42%

For efficiency, industrial large language model (LLM) labelers often pack many candidate labels into one prompt and generate a YES/NO answer for each, even when each decision depends only on the entity and one candidate. Generating these independent decisions as one answer list couples them, spends decoding on the list, and relies on the list being well-formed. In this paper, we present an end-to-end recipe that reformulates the labeler as semantic scoring, fine-tuning it on single-candidate prompts to return each decision as one answer-token score from a prefill-only pass. We also adapt Multiple-Item Scoring (MIS) to score hundreds of candidates in one sequence with attention masked between them. Deployed for profile qualification labeling at LinkedIn, semantic scoring changes at most 0.01% of decisions when candidates are reordered and matches the agreement with reference labels that packed generation reaches only with about 4.7x as many training pairs. The MIS-based stack raises throughput over the deployed stack by 1.24x offline and 1.9x in model serving under a one-second p99 latency target. For offline batch inference, we implement MIS-based scoring as a reusable, model-agnostic Ray Data API proposed for upstream adoption. Model serving uses MIS support contributed to SGLang.

★ MAREval: A Multi-Agent Framework for Evaluating Natural Language Recommendation Explanations

Wed, 28 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

Evaluating natural language explanations in recommender systems is essential for fostering user trust, transparency, and engagement. However, existing evaluation approaches like human evaluations, while accurate, are resource-intensive and impractical at the scale required by modern recommendation platforms. Also, automated methods using single-agent LLMs suffer from prompt sensitivity and inconsistent outputs. To address these challenges, we propose MAREval, a structured multi-agent framework for evaluating recommendation explanations using large language models. MAREval orchestrates (i) a planner agent that uses a novel Chain of Debate (CoD) prompting strategy to coordinate agent roles and enforce logically consistent evaluation plans; (ii) a moderator agent that regulates discussions by mitigating prompt drift; and (iii) an arbitrator agent that aggregates outputs from multiple evaluation rounds and (iv) a Monte Carlo sampling method, improving robustness and alignment with human judgment. We conduct comprehensive evaluations on both public (TopicalChat) and proprietary recommendation datasets, demonstrating that MAREval outperforms state-of-the-art baselines. Comprehensive experiments on a public benchmark and a proprietary e-commerce dataset show that MAREval improves alignment with human judgments over strong singleand multi-agent baselines. Stability analyses indicate substantially lower variability across repeated trials. In a large human-annotation gate, MAREval meets production quality thresholds where prior evaluators fall short, and online A/B testing demonstrates statistically significant improvements in engagement and revenue metrics. These results establish MAREval as a scalable and reliable solution for human-aligned evaluation of recommendation explanations in real-world systems

★ Verifiable Efficiency Limits of Innate LLM Planning in Sliding Puzzles

Yezhou Cheng, Charles Ling — Wed, 28 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

Tool-assisted planning success can reflect external search and verification, leaving a model's unaided planning ability unresolved. We isolate innate planning through complete plans generated in one response without tools or execution feedback. We propose a verifiable framework that decomposes success-weighted plan efficiency as $\mathrm{SPE}=R\times Q\times E$: execution reliability, planning coordination among legal plans, and action economy among completed plans. Trajectory replay and certified optima verify these factors; the joint score includes unsuccessful attempts. We evaluate 8,400 full-plan outputs from seven model conditions on $4\times4$ sliding puzzles, controlling exact horizon and obstacle count in the primary cohort. Models that complete every trial still become less economical at longer horizons, while higher efficiency among completed plans can coexist with poor goal attainment. SPE separates the overall capability hierarchy, while its factors distinguish coordination failure from inefficient completion. The framework makes long-horizon limitations in unaided planning quantitatively verifiable.

★ Your Agent Acts Differently When You're Around: How User Presence Changes Agent Behavior in Negotiation

Jane Pan, You Lyu, Abe Bohan Hou, Valerie Chen, He He — Wed, 28 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

LLM agents increasingly interact with other agents on behalf of users, yet how users in the loop affect their agents' behavior--for better or worse--remains poorly understood. We study this question through the lens of negotiation, where agents must represent users while strategically bargaining with other agents. To isolate the effect of user presence, we develop PRESENT, a controlled evaluation framework that compares negotiation settings with and without an interactive user, while holding the game, objectives, and user preferences fixed across conditions. Across 11.2k runs spanning 7 games and 8 models, we find that user presence significantly changes both outcomes and behavior. Specifically, negotiation parties with a user obtain higher payouts (+13.7% on average) and their agents negotiate more aggressively. Surprisingly, user presence also induces misaligned behavior, both toward counterparties (where its rate roughly doubles) and toward users (where it emerges in 14.3% of games). These misaligned behaviors are uncorrelated with outcomes and persist even when the user sends only templated, neutral messages. These findings expose two blind spots in current evaluation practice: static instructions miss behaviors that emerge only when users are present, and outcome-level metrics cannot detect them.

★ Can LLMs Reliably Grade Olympiad Proofs? A Study of Mathematical Verification with LLMs

Azim Ospanov, Zijin Feng, Ding Ding, Chengwu Liu, Haoli Bai, Jiacheng Sun — Wed, 28 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Hallucination 44%, New model 32%

Large Language Models (LLMs) have made remarkable progress on solving competition-level mathematics, yet their ability to verify natural-language mathematical proofs remains relatively underexplored. This gap is especially pronounced for Olympiad-level proofs, which require meticulous assessment of each claim in a long reasoning chain. Current proof verification thus primarily relies on expert inspection, which is costly to scale. In this paper, we present a systematic study of LLM-based verification in this setting. Specifically, we introduce a human-curated dataset of USAMO 2025 candidate solutions with expert grading, and conduct a broad evaluation of existing LLM verifiers across frontier models. We also propose TROJAN, an iterative self-critique pipeline for generating high-fidelity adversarial proofs at scale. Our comprehensive evaluation covers 20 models spanning seven metrics, two prompt styles, three inference methods, and three prompt templates, showing that state-of-the-art models still leave considerable room for improvement. To narrow this gap, we propose MABGrader, a Multi-Armed Bandit framework that reframes grading as arm selection over the discrete score space. MABGrader outperforms the SOTA ProofGrader by 26.4% in Quadratic Weighted Kappa and reduces grading error by 14%. Code and dataset are available at https://github.com/aziksh-ospanov/TROJAN.

★ Self-Regularized Flow Language Models

Theodoros Kouzelis, Spyros Gidaris, Nikos Komodakis — Wed, 28 Oct 2026 00:00:00 +0000

96% · New model 96%, Efficiency 94%

We introduce SERF (Self-Regularized Flow Language Models), a simple self-supervised framework for improving continuous diffusion language models by exploiting the denoising trajectory itself. While continuous flow-based language models offer a promising alternative to autoregressive generation, their training objectives typically rely on supervision from clean target tokens, leaving the denoising trajectory underexploited as a learning signal. SERF addresses this limitation through noisy-to-clean self-distillation: an exponential moving average (EMA) teacher receives a cleaner version of the student's noisy input along the same interpolation path and provides additional supervision. We investigate two variants: SERF-L, which aligns the student's intermediate representations with deeper teacher representations, and SERF-P, which aligns their output token distributions, either at the student's output or at an early student layer. All variants augment the standard training objective without requiring external pretrained models or modifying the inference procedure. Experiments across three continuous language-model paradigms (one-hot FLM, latent ELF, and hyper-spherical S-FLM) demonstrate substantial improvements in training efficiency, generation quality, and few-step generation. In particular, SERF achieves up to 7.7x wall-clock training acceleration on OpenWebText, improves MAUVE by 24.3% and GSM8K accuracy by up to 28.8%, and matches baseline performance with up to 64x fewer sampling steps. These results establish noisy-to-clean self-distillation as an effective approach to improving continuous diffusion language models.

★ Channel Selection for Mixture-of-Experts Decoding Needs No Calibration Data

Geunhyoung Kim, Sung Ju Hwang — Wed, 28 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, Evaluation 37%

Mixture-of-Experts (MoE) models are bottlenecked by loading routed-expert weights from memory during interactive decoding at small batch sizes. Because most intermediate channels in a routed expert contribute little to any single token, channel selectors predict which channels matter and read only their weights, typically keeping channels above a threshold calibrated on offline data. We show that channel selection needs no calibration data. First, a calibrated threshold does not fix how much a selector reads, because the number of channels passing it varies across inputs; Prox with thresholds calibrated for an 18% read budget actually reads only 12.9% on average across our benchmarks. Second, top-k keeps each token's k highest-scoring channels and needs no calibration data. At the calibrated budget, a selector's top-k version matches or beats its threshold version on the same score in 40 of 43 comparisons at 13% and 18% budgets, including all 23 on generation tasks. The top-k versions also avoid the repetition-loop collapse that calibrated thresholds cause on Gemma 4. Calibrating a threshold thus provides neither of the two things selectors use it for: a known read budget and a consistent accuracy gain at that budget. We furthermore introduce a reader built entirely from the model's weights, which screens channels with a ternary sketch and recovers the skipped ones with a low-rank map. Across three post-trained MoE models, our reader retains 98.1% of full-model accuracy at 17.8% of routed-expert reads, with Triton kernels running MoE blocks 1.60 to 1.90x faster than optimized vLLM at batch one and up to 2.69x at larger batches on the same three models.

★ CRISP: Co-designing Pruning and Scheduling for Efficient MoE Inference on Edge Servers

Congwei Zhang, Youming Tao, Yifei Zou, Yu Yang, Ruirui Zhang, Xiaoyu Zhang — Wed, 28 Oct 2026 00:00:00 +0000

96% · Efficiency 96%

Mixture-of-Experts (MoE) inference on memory-constrained edge platforms is often dominated by expert-weight movement rather than computation. Although each token activates only a small subset of experts, routing is input-dependent, so all experts must remain accessible and memory movement dominates latency. Existing approaches such as uniform pruning and expert dropping overlook two properties of MoE execution: expert combinations are highly skewed in frequency, and experts contribute asymmetrically within each combination. We present CRISP, a combination-aware framework that co-designs pruning and runtime scheduling for efficient MoE serving on memory-constrained edge platforms. CRISP reorders neurons by importance so one stored expert can support multiple execution widths, assigns differentiated execution to hot, warm, and cold combinations based on frequency and expert role, and exploits heterogeneous widths to overlap expert loading with computation. Across Mixtral-8x7B and DeepSeek-MoE-16B, CRISP improves throughput by up to $2.1\times$ at the same memory budget while limiting accuracy loss to 3.6\% relative to full-width inference. Our results highlight that efficient edge MoE serving requires jointly optimizing which expert capacity is preserved and how expert execution is scheduled.

★ Can LLMs Uncover Hidden Physical Laws through Free Exploration?

Zhen Yang, Wen Wang, Xander Xu, Ying-Cong Chen — Wed, 28 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Agents 81%

Scientific discovery often begins with an unexpected observation that prompts a new research question. In a world with altered physical laws, can large language models (LLMs) notice unexpected behavior, such as heavier objects falling more slowly, and discover the law through experiments without being told what to investigate? To evaluate this ability, we introduce LongChang, an automated framework that creates simulated worlds with altered physics, where one or more laws differ from ordinary physics. The evaluated LLMs act as explorers, manipulating objects and building tools to investigate their own questions without research tasks or hints about these laws. Separate LLM agents generate and review scenes, review tool designs, and evaluate reports for correctness and experimental support using the altered physical laws and exploration records. These records guide scene evolution for further exploration, allowing the full cycle to run without human intervention. Among 12 LLMs, GPT-6-Astra achieves a discovery rate of 16.7%, while the other 11 score 0%. The exploration records reveal uninvestigated anomalies and only partial identification of the altered physical laws.

★ Distributed LLM Inference via Communication-Efficient Speculative Decoding

Joohyun Shin, Byeonghyeon Park, Keumseo Ryum, Joonhyuk Kang — Wed, 28 Oct 2026 00:00:00 +0000

95% · Efficiency 95%

Distributed speculative decoding has emerged as a promising approach for device--server collaborative large language model (LLM) inference, but its decoding speedup is often limited by the vocabulary-scale communication overhead. In existing distributed speculative decoding methods, exact rejection sampling requires transmitting full token distributions over the LLM vocabulary between the device and the server, which becomes costly in bandwidth-constrained environments. To address this issue, we propose DISCO, a communication-efficient framework for distributed speculative decoding, that eliminates vocabulary-scale distribution transfer while preserving the exact target distribution. DISCO achieves this by introducing a server-side replica of the draft small language model (SLM), which enables the server to locally obtain the draft-side distribution required for rejection sampling. As a result, DISCO exchanges only drafted token indices and scalar probability values between the device and the server. Experimental results on multiple datasets show that DISCO consistently achieves higher decoding speedup than existing distributed speculative decoding methods, particularly under stringent bandwidth constraints.

★ OA-RAG: Ontology-based Agentic RAG for Faithful LLM Reasoning

Hui Yang, Jiaoyan Chen, Yongsheng Gao — Wed, 28 Oct 2026 00:00:00 +0000

95% · Hallucination 95%, Evaluation 82%

Large language models (LLMs) perform well across many real-world tasks but remain prone to hallucination, limiting their reliability in domains such as biomedicine and healthcare, where answers must be grounded in established knowledge. Retrieval-augmented generation (RAG) mitigates this problem by incorporating external information, but often overlooks logical relationships and relies heavily on the LLM's own reasoning, which can compromise answer faithfulness. We propose OA-RAG, an agent that integrates formal reasoning with LLM reasoning, drawing on both the logical structure and informal annotations of domain ontologies. For each answer, OA-RAG reports the ontology observations used as evidence and a model-assessed grounding level (entailed, supported, related, or unrelated) that characterises the strength of that evidence. Across benchmarks derived from three large-scale ontologies and six real-world medical question-answering tasks, OA-RAG achieves the best overall performance among the evaluated ontology-fine-tuned and popular RAG methods, with particularly strong results on reasoning-intensive tasks. The reported grounding levels also offer a useful indication of answer reliability, with stronger grounding generally associated with higher accuracy. Finally, our results demonstrate the value of ontologies as knowledge sources: OA-RAG is competitive with, and sometimes outperforms, domain-specific RAG systems that draw on sources such as textbooks.

★ HiLo-RD: Cross-Budget Self-Distillation for Masked Diffusion Language Models

Xiang Li, Chengting Yu, Jungang Li, Yilun Zheng, Junzhuo Li, Xingyao Zhang — Wed, 28 Oct 2026 00:00:00 +0000

95% · New model 95%, Efficiency 84%

We propose HiLo-RD (High-to-Low Budget Representation Distillation) to improve masked diffusion language models (MDLMs) under a limited generation budget. Unlike autoregressive models, MDLMs choose both what to generate and where to commit it. This creates two coupled challenges: predicting useful content before reasoning is complete, and deciding when to fix answer tokens that shape later predictions. We identify budget-induced premature commitment: even when a correct solution fits, the model can fix answer fragments near the window boundary before resolving intermediate steps. These fragments become context for later predictions and redirect subsequent reasoning. Existing post-training methods do not directly transfer the stronger trajectories that the same model follows with a larger window. HiLo-RD addresses this gap by (1) pairing large-window teacher states with cropped small-window views and (2) matching their hidden representations at committed positions. This supervision guides both token prediction and commitment order without an external teacher. Across mathematics benchmarks, HiLo-RD improves low-budget accuracy and shortens responses; these gains extend to code generation without code-domain training. On GSM8K, accuracy rises by 9.4 percentage points as mean readable length falls by about 25%. Attention diagnostics and fixed-budget order interventions support the proposed mechanism. Together, these results show that large-window reasoning can improve small-window decoding.

★ TDLM: Adaptive Temporal Integration for Diffusion Language Models

Bohang Sun, Max Zhu, Francesco Caso, Christopher Irwin, Jialin Yu, Pietro Li\`o — Wed, 28 Oct 2026 00:00:00 +0000

95% · New model 95%, Efficiency 95%

Masked diffusion language models repeatedly compute representations for unresolved tokens, but conventional denoisers do not explicitly retain these representations for reuse across steps. We introduce Temporal Self-Conditioning Diffusion Language Model (TDLM), which lets each unresolved position attend to its own history of hidden representations within a frozen denoiser. This gives later predictions access to computation already performed. Training uses self-generated trajectories and a simple cross-entropy objective for tokens revealed between consecutively sampled decoding steps, while causal temporal attention permits parallel evaluation during training. On LLaDA, the temporal modules add only approximately 0.84% trainable parameters. Across LLaDA and Dream, the main comparison shows up to 7.15$\times$ faster generation than the corresponding original diffusion language model, with competitive answer quality on the evaluated tasks. These results motivate temporal memory as a practical way to make more effective use of repeated denoising computation.

★ Live-OPSD: On-Policy Self-Distillation for Proactive Streaming Large Video Models

Kaixuan Du, Xin Wan, Dai Guan, Ming Chen, Xiaoxi Jiang, Guanjun Jiang — Wed, 28 Oct 2026 00:00:00 +0000

95% · New model 95%, Evaluation 38%

Streaming multimodal assistants must make response decisions while events are still unfolding. They need to recognize when the available evidence warrants a response, produce an appropriate continuation, and remain silent otherwise. Supervised fine-tuning learns from recorded assistant turns, but does not directly train on the response prefixes produced by the evolving policy itself. We introduce Live-OPSD, an on-policy self-distillation method that uses privileged hints available only during training to supervise the policy on its own outputs. The policy first generates a continuation from the streaming context alone. The same model then re-evaluates that continuation with a hint describing the relevant evidence, and the resulting distributions guide the unhinted policy. On Qwen3-Omni-30B-A3B, Live-OPSD attains an average score of 59.64 on LiveProBench, 5.26 points above the strongest continued-SFT baseline, and stays competitive on OVO-Bench (65.84). These results indicate that guiding the policy on its own outputs with privileged hints improves both when and how a streaming assistant responds.

★ Metacognitive Decoding: Efficient Inference for Diffusion Language Agents

Mingyu Zhao, Shuhang Lin, Ligong Han, Xiaojing Yu, Zhenting Wang, Yifei Pang — Wed, 28 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, New model 70%, Agents 52%

Diffusion language models offer a promising path to efficient agents through parallel token generation. However, efficient interaction requires adapting both where computation is spent and how much text is generated. We introduce a metacognitive controller that guides diffusion generation according to the needs of the current response and the agent's interaction with the environment. It learns when to finalize tokens and how far generation should extend, reducing overall computation within responses. Our controller, a lightweight plug-in for the diffusion backbone, is trained on agent trajectories to balance correctness and generation cost. This approach connects the agent's generation process to the interaction the agent serves. Experiments across five benchmarks demonstrate that our approach reduces inference FLOPs by 20.6% on average compared with the most efficient baseline, while maintaining competitive task accuracy. Ablation studies further validate the effectiveness of our approach.

How Much Does the Hint Matter? Scope Conditions for Hindsight Self-Distillation in Financial Reasoning

Johnny Yu, Cyril Gorlla — Wed, 28 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 90%, Efficiency 75%

Financial institutions operating under data-residency constraints are restricted to open-weight models small enough to locally host, which trail frontier systems on multi-step quantitative reasoning. On-policy distillation narrows such gaps by supervising a student on its own rollouts, but traditional OPD requires a strong external teacher, and on-policy self-distillation (OPSD) conditions on a full reference solution. We ask how much of that signal is necessary. We present Fin-SD, a localized hindsight self-distillation recipe for multi-step reasoning. Given a problem the student fails, the student itself identifies the first incorrect reasoning step and writes a short critique of that step. We truncate the student's trajectory immediately before the error (aiming to obtain a correct prefix), condition a frozen copy of the student on the same prefix plus the hint, and distill that hint-conditioned teacher into the student over a short rollout under reverse KL. No external model contributes a gradient. We apply Fin-SD to GPT-OSS-20B on a custom multi-step financial reasoning benchmark that requires deriving formula inputs from realistic scenarios via accounting conventions. We find: (i) replacing the critique with a generic-feedback control matches Fin-SD's entire 3.03 pp accuracy gain over the base model on the FinanceReasoning Hard subset, suggesting the gain comes from where supervision is applied, not hint content; (ii) the model writing the hint does not need to be stronger than the model receiving it: hints from same-family GPT-OSS-120B, at roughly 6x the parameters, yield no materially greater benefits than self-written hints; and (iii) Fin-SD matches on-policy self-distillation from full reference solutions on accuracy while generating 9.5% fewer tokens. With Fin-SD, general reasoning on MMLU-Pro and a fourteen-metric compliance audit on FinTrust show no material regression.

Proposing Better Rollouts for On-Policy Distillation

Sirui Chen, Yunzhe Qi, Haobo Xu, Jiaru Zou, Mengting Ai, Zihao Li — Wed, 28 Oct 2026 00:00:00 +0000

94% · New model 94%, Efficiency 65%, Evaluation 39%

On-policy distillation (OPD) improves language-model post-training by letting the student learn from its own sampled trajectories under teacher supervision, aligning the training process with inference-time behavior. However, the rollout policy in OPD is usually treated as fixed: each round samples directly from the current student. We show that this choice can create a compounding teacher--student mismatch, because it requires the teacher to provide effective supervision on whichever trajectories the student samples, including ones less reachable by the teacher itself. On such trajectories, teacher supervision can become noisy or even misleading, leading to undesirable student patterns such as repetition. Moreover, since the updated student in turn defines the rollout policy of subsequent rounds, these patterns persist and potentially amplify over training. This failure mode suggests that improving OPD requires treating the rollout policy itself as an optimization target, rather than only modifying the distillation objective. We propose Mixture Proposal On-Policy Distillation (MPOPD), which derives a student-anchored, teacher-guided rollout distribution from a local trust-region objective based on relative teacher support. The resulting target is a geometric mixture of the student and teacher policies, and we introduce a proposal policy that approximates this mixture, enabling rollout generation in a single forward pass. Experiments on 16 math, knowledge, and medical reasoning benchmarks show consistent gains over OPD and recent variants. On math reasoning, our method improves average accuracy by 6.91 over OPD on Qwen3-4B-Base while reducing average generation length by 41%, suggesting that rollout policy design provides an effective way to improve OPD.

SetGRPO: Calibrating Grader-Induced Policy Optimization

Jiale Huang, Zixu Li, Zhiwei Chen, Zhiheng Fu, Jinhe Bi, Yupeng Hu — Wed, 28 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 53%

Rubric-based reinforcement learning uses fine-grained grader feedback to train large language models on open-ended tasks, but grader randomness can alter GRPO updates in ways that score discrepancies alone cannot characterize. Because GRPO jointly normalizes rewards within each response group and combines the resulting training coefficients with response-level gradients, a scoring discrepancy can shift the coefficients of the entire group and induce policy-dependent update changes. Selective execution further affects the direction-reversal risk of retained updates. To address this, we propose Set-Valued Group Relative Policy Optimization (SetGRPO). It comprises three modules: (a) Joint Reward Set Calibration constructs a joint uncertainty set for the training coefficients; (b) Geometry-Guided Safe Recovery maps candidate coefficients into the current policy's local update geometry and determines the largest safe recovery ratio for each response group; and (c) Selective Risk-Gated Optimization selects a global threshold through independent risk calibration to control the conditional direction-reversal risk of accepted updates while preserving the original GRPO objective. Under exchangeability, SetGRPO provides finite-sample marginal coverage for the joint coefficient set and simultaneous risk control across candidate thresholds. On ten open-ended benchmarks spanning three domains, SetGRPO achieves the highest macro average among all compared training methods on both Qwen3.5 and Llama3.2 model families.

When All Rollouts Fail: Policy-Compatible Joint Decoding for Hard-Problem Reasoning

Zehao Wang, Hanzhu Chen, Runquan Gui, Zhe Wang, Lin Yang — Wed, 28 Oct 2026 00:00:00 +0000

94% · New model 94%

Reinforcement Learning with Verifiable Rewards (RLVR) derives its learning signal from binary rewards via on-policy exploration and becomes a powerful paradigm for improving the reasoning of large language models. Yet it faces a learning cliff on problems beyond model's current ability: when every rollout fails, the samples reveal nothing about a correct solution, and group-relative estimators (e.g., GRPO, RLOO) collapse the update to zero advantage. Expert trajectories offer a remedy, but they are incompatible with the current policy and often lie in the student's low-probability regions, so the surviving updates demand large parameter shifts and decay into surface-level imitation rather than transferable reasoning. We argue that a useful learning signal on hard problems must be both verifiably correct and sufficiently compatible with the current policy. We propose Policy-Compatible Reinforcement Learning (PICO), which performs distribution-constrained joint decoding: expert data guides decoding at the token level while preserving the student's relative token probabilities, and a dynamic top-k mechanism confines the sampling distribution under a finite divergence budget, yielding trajectories that are verified and policy-compatible by construction. These trajectories are optimized jointly with on-policy rollouts under a mixed-behavior objective that mitigates their distribution mismatch via importance sampling. Experiments across reasoning diverse datasets and agentic benchmarks demonstrate that PICO consistently outperforms standard RLVR and prior off-policy baselines.

GUI-JEPA: Candidate-Conditioned JEPA Latents as a Fast Action Policy for Mobile GUI Agents

Bingchao Wang, Yongcheng Jing, Shunyu Liu, Chengyu Wang, Jun Huang, Jie Yang — Wed, 28 Oct 2026 00:00:00 +0000

94% · Efficiency 94%, New model 61%

Mobile GUI agents are typically fine-tuned 7B-scale vision-language models (VLMs) that generate each action as text. However, these VLMs decode every action token by token, and even agents that score candidate actions keep a large language model or VLM in the loop, making them slow and memory-hungry. To address these issues, we propose GUI-JEPA, to our knowledge the first decoder-free GUI agent policy built on a Joint-Embedding Predictive Architecture (JEPA), which selects actions in the latent space of a JEPA encoder pretrained on unlabelled GUI screenshots. Its action- and region-conditioned latent predictor turns the screen latent into one latent per candidate action, letting lightweight heads score all candidates against the instruction in a single batched pass, without decoding a token. We further show that the scorer relies on this latent pathway and that the predictor is trained for the decision itself: a randomly initialised predictor does just as well as one pretrained on screen dynamics. On AndroidControl, GUI-JEPA is 2.8-4.0x faster on the same GPU than a Qwen2.5-VL-7B agent we fine-tune with the same inputs, while reaching 86-92% of its in-distribution accuracy with 6% of its parameters. Code is available at https://github.com/bcwang-sjtu/GUI-JEPA

SimplexLoRA: Dynamic Rank Updating via Alternating Minimization with Simplex Projection

Wed, 28 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 35%, Efficiency 33%

The success of large language models in real-world applications relies on parameter-efficient fine-tuning methods such as Low-Rank Adaptation (LoRA), which inserts compact, trainable adapters into selected layers. In this paper, we propose SimplexLoRA, a LoRA variant that adaptively adjusts adapter ranks during fine-tuning based on their learned importance. Unlike standard LoRA, which employs a fixed rank across all layers, SimplexLoRA optimizes per-adapter importance weights on the $n$-scaled simplex during a short warm-up and then freezes them, dynamically allocating capacity where it is most beneficial without increasing the overall parameter budget or inference cost. We evaluate SimplexLoRA across 12 language understanding, summarization, and reasoning tasks using DeBERTaV3, BART-large, Llama 3.1 8B, and Qwen3-8B. Our results show that SimplexLoRA consistently matches or outperforms standard LoRA and recent adaptive baselines, with the largest gains at low ranks. Our code is available at https://github.com/brain-lab-research/SimplexLoRA.

Decision Models as Judges for Hallucination Detection

Haozheng Liu, Zhaoqian Xue, Zhongyin Wang, Junting Mu, Zihui Ma, Shuya Feng — Wed, 28 Oct 2026 00:00:00 +0000

94% · Evaluation 94%, Hallucination 90%, Efficiency 85%

Large language models are widely used as judges of model outputs (LLM-as-a-Judge), but they are costly and can hallucinate or be overconfident. Decision models are a potential alternative. They return a predefined verdict with a probability for each label instead of free-form text and are much cheaper and faster. We test this promise on hallucination detection. We build binary benchmarks for five hallucination types (9,852 items from six datasets) and compare five decision models with the generative judge gpt-6-sol. Performance depends strongly on the type: the best decision model matches gpt-6-sol in balanced accuracy on context inconsistency and comes within 3 percentage points on factual errors, but trails it by 4-8 points on logical and instruction inconsistency. Decision-model probabilities are overconfident, though less so than the GPT-6 models, and separate correct from incorrect decisions only moderately. We also find that a confidence cascade of decision models improves on the best single model by 1.9-3.3 points on three types and matches or exceeds gpt-6-sol at 1-3% of its cost. Overall, decision models are a practical low-cost judge for some hallucination types but need to be validated before use.

MuonGS: Pretraining LLMs with Relaxed Weight-Norm Control

Athanasios Glentis, Dawei Li, Mingyi Hong — Wed, 28 Oct 2026 00:00:00 +0000

94% · New model 94%

Weight decay and explicit weight-norm constraints improve optimizer performance in language model pretraining, yet a unified understanding of why such control helps and how much is needed remains incomplete. We analyze norm control in Muon over the entire training trajectory using the \emph{accumulated effective learning rate}, $C_T=\sum_{t<T}\eta_t/\|W_t\|_F$, where $\eta_t$ is the nominal learning rate and $\eta_t/\|W_t\|_F$ is the effective learning rate relative to the weight norm. For bounded update directions, accumulation over any training interval bounds relative weight displacement. We show that vanilla Muon can accumulate only logarithmically, potentially limiting exploration and helping explain its slower empirical progress, whereas MuonW and MuonH guarantee at least linear accumulation under the stated conditions. Going beyond bounded-norm control, we show that sufficiently slow norm growth guarantees faster accumulation than the logarithmic worst case, opening up less restrictive optimizer designs. Based on this insight, we propose MuonGS, which uses Gram--Schmidt (GS) projection to remove the radial component of each Muon update without fixing the weight norm. With bounded update directions and a constant nominal learning rate, we prove that the weight norm grows at most proportionally to $\sqrt{t}$ after $t$ steps. Consequently, MuonGS guarantees $C_T=\Omega(\sqrt{T})$, with accumulation sustained even in later training. Further, for smooth scale-invariant objectives we establish its $O(T^{-1/2})$ stationarity guarantee. In pretraining experiments with 160M- and 400M-parameter dense models as well as a 135M active- and 1B total- parameter MoE model, MuonGS achieves lower validation perplexity than MuonH at long training horizons. In particular, for the dense model it achieves an estimated token-equivalent speedup up to about $15\%$ at matched validation loss to MuonH.

DuplexTune: Coupled-State Replay for Memory-Efficient Mixture-of-Experts Adaptation

Paul Raj, Simon Caldwell, Daniel Whitmore, Mutlu Mete — Wed, 28 Oct 2026 00:00:00 +0000

93% · New model 93%, Efficiency 87%, Evaluation 36%

Sparse expert execution does not remove the stored states needed to adapt a mixture-of-experts language model. We present DuplexTune, a framework for replaying coupled hidden states through pretrained attention and expert modules. Trainable width bridges connect two residual streams to the backbone. Bridge warm-up precedes backbone adaptation, with router weights kept frozen. We analyze the implicit state-recovery equation and identify the convergence and replay conditions needed for reconstruction. On Qwen1.5-MoE-A2.7B, the single-H800 evaluation reports 38.8 GB peak memory, versus 64.8 GB for supervised fine-tuning with checkpointing, and throughput of 25.1 versus 19.9 samples/s. DuplexTune achieves 67.2% MMLU, 75.6% GSM8K, and 7.69 MT-Bench, with a weaker multilingual score than the checkpointed baseline. The results characterize a configuration-dependent tradeoff among state storage, throughput, and adaptation quality.

CohortFlow: A Steerable Clinical Cohort Curation Agent with Localized Semantic Correction

Neil Chen, Adrian Serapio, Hanxue Gu, Evan Lee, Kang Wang, Yang Yang — Wed, 28 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 71%

Language model agents for clinical cohort curation should accommodate evolving eligibility criteria without repeatedly processing unchanged requirements. We introduce CohortFlow, a steerable agent that represents natural language corrections as localized updates to a structured cohort specification and reconciles them with a persistent execution graph. Explicitly recorded semantic dependencies enable affected computations to be selectively invalidated while preserving valid prior results. On 100 scenarios derived from clinical-trial eligibility criteria, comprising 126 correction turns, localized steering achieves 77.0% requirement micro-F1 versus 76.4% for full regeneration using the same 8B model, a small numerical advantage that is not statistically significant ($p=0.493$). It reduces mean correction latency by 64.8% after completion and 44.3% during execution, with 38.1% fewer model calls. In a separate proof-of-concept evaluation over MIMIC, all 13 cohort states execute successfully, while correction trajectories demonstrate selective preservation of unaffected computations and recomputation of changed or dependent nodes. Together, these results show that localized semantic steering can substantially reduce correction latency and model calls while maintaining strong requirement-level accuracy.

Base Models Can Detect Hidden Intentions that Prompted Classifiers Miss

Oz Amram — Wed, 28 Oct 2026 00:00:00 +0000

93% · Evaluation 93%

The Persona Selection Model (PSM) posits that post-training selects a persona the model learned during pre-training. Work on subliminal learning has shown that models can impart covert signals (like "love of owls") into semantically unrelated outputs like strings of numbers, and that students trained on these outputs inherit those traits. In this work I show that by evaluating the likelihood of such outputs under the base pre-trained model with an appropriate header, a method I call persona-header likelihood evaluation (PHLE), one can decode these subliminal signals and classify between them. For example, PHLE distinguishes number sequences generated by a post-trained model prompted with "You love owls" from those prompted with "You love eagles", and distinguishes answers from a model with a secret desire to harm the user, filtered to remove any noticeable misalignment, from those of a model with no system prompt. In both cases prompted classifiers, including the generating model itself, remain near chance. The signal partially persists across model families, with the Qwen base model reading signals generated by Olmo and vice versa. This is evidence for a non-trivial prediction of the PSM, and the method may be practically useful for auditing the hidden objectives of models.

From Solver Trajectories to Teaching Trajectories: Cognition-Aligned Reasoning Distillation

Yashuo Luo, Qianren Mao, Ziqi Qin, Junnan Liu, Haitao Li, Qili Zhang — Wed, 28 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 34%

Knowledge distillation from large reasoning models (LRMs) often uses raw teacher trajectories as supervision. However, this practice creates a mismatch between the goal of distillation and the nature of teacher trajectories. Advanced LRMs typically improve their reasoning capabilities through reward-driven post-training, such as RLHF. This training optimizes them as strong solvers, rather than as teachers that produce reasoning processes smaller models can easily learn. Since smaller models have different "cognitive capacities" compared to their larger teachers, directly imitating raw teacher trajectories can sometimes be ineffective and often requires a substantial number of high-quality samples. To bridge this mismatch, we propose Cognition-Aligned Reasoning Distillation (CARD), an RL-based framework that transforms solver trajectories into teaching trajectories aligned with the student's cognition. Avoiding expensive resampling, our framework directly transforms existing trajectories. We optimize CARD with two reward signals: (1) a Distribution Alignment Reward that matches the student's predictive distribution, and (2) a Logical Coherence Reward that reduces misleading failed reasoning steps. Comprehensive evaluations across multiple student models and settings demonstrate that CARD outperforms distillation from raw teacher trajectories in in-domain settings while achieving substantial gains in out-of-distribution performance. Code is available at https://github.com/LiteSSLHub/CARD.

Guided-RLVR: Guided Exploration in Reinforcement Learning with Verifiable Rewards

Bo Yang, Zishuai Zhang, Weihua Li, Qinnan Zhang, Sijia Wen, Hainan Zhang — Wed, 28 Oct 2026 00:00:00 +0000

93% · New model 93%

Reinforcement learning with verifiable rewards (RLVR) has shown strong empirical gains in improving the reasoning ability of large language models (LLMs) on tasks such as mathematics and programming. However, recent studies have shown that RLVR rollouts often collapse onto a narrow set of reasoning trajectories, limiting the explored reasoning space. Motivated by findings that thinking tokens at critical steps serve as critical structural elements in LLM reasoning pathways, we propose Guided-RLVR, a novel approach that augments RLVR with a lightweight controller. The controller strategically injects semantically meaningful steering tokens into the sampling distribution, encouraging the model to approach problems from multiple reasoning perspectives and thereby expanding the exploration space. Experiments on mathematical reasoning and code generation benchmarks show that Guided-RLVR consistently outperforms existing RLVR methods while producing more diverse reasoning trajectories.

Does Your Generalist Skill Optimizer Really Incentivize Agentic Tasks? Rethinking Skill Optimization via Bipartite Attribution

Jiawei Zhou, Bo Peng, Yuxuan Huang, Ming Zhang, Youbang Sun, Xiang Wang — Wed, 28 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, Agents 61%, New model 41%

Skill optimizers adapt an agent to long-horizon tasks, such as repairing a terminal environment, by rewriting its skill, a natural-language procedure in its context, and keeping the edits that raise success on a small dev set. On such tasks success is noisy, so a gain on a small dev set does not show that an edit helps, and a success score does not show what the edit costs. We introduce BiPA, which replays an optimizer's acceptance rule on candidates identical to its seed to measure what selection alone reports, and records each episode as a bipartite graph of turns and the text they read to trace extra input tokens to the behavior behind them. Across three optimizers, three agentic benchmarks and two models, the learned skills succeed about as often as their hand-written seed, 0.1 points higher on average. Their dev gains vanish on test: keeping the best of several candidates on at most 30 tasks reports gains this large even when none changed. They read about 16% more input tokens: the optimizers almost always add rules, half to re-check work and some to act one call at a time, and the agent follows them as extra turns, each re-sending the whole conversation. Four principles follow: accept an edit only if its gain exceeds what selection alone reports; charge it for the turns it adds, not for its length; let the editor delete and merge rules; show it what earlier rules cost. Code is available at https://github.com/valleysprings/BiPA.git.

Concordat: Typed Sessions for LLM Agents

Milad Sikaroudi — Wed, 28 Oct 2026 00:00:00 +0000

93% · Agents 93%, Evaluation 57%

Agents built on large language models are increasingly allowed to take consequential actions: cancelling bookings, issuing refunds, changing accounts. The rules that govern those actions, such as "obtain the customer's explicit confirmation before changing anything", are usually stated in the prompt and followed only as reliably as the model chooses to follow them. We present Concordat, a Python library that treats an agent as an object whose interface is a finite-state protocol, in the spirit of typestate and session types, and enforces that protocol on the model at run time. On every turn the model is offered only the tools that are legal in the current protocol state, calls outside that set are rejected before they have side effects, decisions reserved for humans can only be requested, and token and tool-call budgets are charged and split linearly across handoffs. Protocols are checked when they are defined, and a subclass of an agent is rejected if it could bypass its parent's protocol, a behavioural form of Liskov substitution. We evaluate a Concordat agent against the standard agent of tau-bench on its 50 airline tasks, adding simulated customers who pressure the agent to skip confirmation. The standard agent made changes without the required "yes" in 13 of 50 tasks with gpt-4.1 and in 29 of 128 task-runs with Claude Haiku 5.5; the Concordat agent made none, with no significant difference in task success. tau-bench's own reward did not detect most of these violations. Concordat is open source.

What Improves During Pretraining? Tracking Knowledge Acquisition with Point-in-Time Models

Chang Su, Bin Wu, Emine Yilmaz, Yao Lu — Wed, 28 Oct 2026 00:00:00 +0000

93% · Evaluation 93%

Language models pretrained on web corpora collected over multiple years learn both general and domain-specific knowledge. How these different kinds of knowledge are learned as pretraining progresses remains unclear. Recent point-in-time models trained on web snapshots from different years provide a way to probe how knowledge prevalence varies over time. Using these models, we propose a temporal variability score (TVS), defined as the coefficient of variation of an item's negative log-likelihood across models. A higher score suggests that the knowledge may be less consistently represented across years of web text. We evaluate six benchmarks covering diverse topics using six models from multiple open-source model families. We find that low- and high-TVS items are learned similarly early in training, but high-TVS items improve faster later on. At the end of the pretraining run, high-TVS items outperform low-TVS items in 34 of 36 model-benchmark settings, with an average advantage of 7.3 points of chance-normalized accuracy. This pattern is most robust on knowledge-intensive tasks, even when comparing items of similar difficulty. Our findings reveal differences in how knowledge is learned over the course of pretraining, with temporal variability providing a diagnostic of these learning dynamics beyond aggregate benchmark scores.

Balanced State Subspaces for Mamba-2 State Compression

Shang Tingyu, Cao Wenfei — Wed, 28 Oct 2026 00:00:00 +0000

92% · Efficiency 92%, Evaluation 39%

Mamba-2 compresses all context into a fixed-size recurrent state that dominates decoding memory and bandwidth. Existing compressors quantize state values or prune state coordinates, scoring each channel independently in the native basis. We examine trained Mamba-2 states and find that energy is almost rank-one, but the output does not read the energetic directions: energy concentrates in a bias-like direction accumulated by the recurrence, while the content the output reads lives in a low-dimensional subspace that is not aligned with the state's coordinates. Even output-aware scores are computed coordinate by coordinate, so they miss the directions that determine the input--output map. Compression should therefore target the write--read subspace rather than individual coordinates. We propose \method{}, a post-training transform that instantiates balanced truncation with full controllability and observability Gramians and applies the rotation exactly via Mamba-2's scalar-identity transition. Because the useful rank varies widely across layers, we further introduce HARP, which allocates ranks per layer from the Hankel spectrum and protects retrieval-sensitive layers, improving both perplexity and retrieval over uniform budgets. Calibration takes a single forward pass, an order of magnitude faster than GHOST's released protocol, with lower peak memory. With HARP, \method{} halves the state dimension at an average WikiText-2 perplexity increase of $0.3$ points ($1.8\%$) across 130M--2.7B, versus $1.6$ ($9.3\%$) for GHOST, and the gap widens at 25\% retention.

CAMR: Compatibility-Aware Memory Replay for Continual Learning in Large Language Models

Jingpu Yang, Tianhao Li, Fengxian Ji, Yufeng Wang, Shuxin Zheng, Xiaoqing Zhang — Wed, 28 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 48%, Efficiency 43%

Continual learning requires large language models to acquire new knowledge and capabilities while retaining what they have learned. Fine-tuning and replay support acquisition and retention, but deciding whether incoming examples need supervision and which historical examples to revisit remains a challenge. We propose Compatibility-Aware Memory Replay (CAMR), an adaptive framework that jointly controls current-example supervision and historical replay. CAMR combines compatibility checks with a learned memory gate to guide an online policy over three actions: learning incoming examples with general replay, jointly reinforcing compatible related examples, and skipping direct supervision of verified mastered duplicates. The policy learns from acquisition, retention, and training-cost feedback under a bounded per-arrival replay budget. We evaluate CAMR with Qwen3-8B-Base on continual factual updating and cross-task and cross-domain learning. On WikiBigEdit, CAMR consistently leads evaluated methods in Original Match from 3K to 100K facts, reaching 93.66% after just 3K facts and surpassing Full-history Replay (LoRA) after 100K facts. It also achieves the highest final scores on all four selected TRACE tasks and maintains or improves earlier-task performance relative to post-acquisition levels in a sequence spanning financial question answering, mathematical reasoning, code generation, and summarization.

API-TAIL: Detecting Model Substitution in LLM APIs through Temperature-Aligned Likelihood Tails

Haocheng Mei, Yuchuan Tian, Ye Yuan, Chao Xu, Allen Chen, Kai Han — Wed, 28 Oct 2026 00:00:00 +0000

92% · Evaluation 92%

Commercial language model APIs return only text, allowing a provider to cut costs by quietly routing a fraction of user requests to a cheaper substitute model while billing for the advertised one. Detecting such partial substitution is challenging because honest providers also alter the output distribution through undocumented decoding and serving choices, such as temperature scaling, sampling truncation, repetition and presence penalties, and default system prompts, forcing an audit to test against a composite null of legitimate configurations rather than a single reference. Our key insight is that these legitimate variations primarily shift response typicality across the entire batch, whereas partial substitution injects individual responses that remain anomalous under the claimed model. Building on this, we introduce API-TAIL, a typicality audit that first absorbs batch-level shifts by aligning a single effective temperature to center standardized surprisal scores across the audited batch, and then flags the endpoint if the number of remaining tail responses exceeds a threshold calibrated across legitimate configurations. Evaluated on real-world commercial APIs, API-TAIL reliably controls false alarms across diverse honest providers while detecting even low-rate partial substitution with a small query budget, substantially outperforming existing rank- and MMD-based audits that lose detection power under the same composite calibration.

Prospector: learning a compact graph traversal policy for scientific reasoning

Haiqian Yang, Markus J. Buehler — Wed, 28 Oct 2026 00:00:00 +0000

92% · Efficiency 92%, New model 87%, Agents 78%

Scientific reasoning in specialized engineering fields depends on knowledge scattered across the literature, often proprietary and absent from frontier LLM training data. In such domains, frontier models tend to give generic answers. Knowledge graphs (KGs) can ground LLM reasoning. Because a graph is too large to provide in full, typical LLM+KG systems preselect a portion, deciding relevance before reasoning begins. Here we train a compact model, Prospector, to reason over a domain knowledge graph by traversing it step by step, seeking evidence in light of what it has already found. Prospector calls a small set of tools one at a time to progressively build its context, and answers with a chain of evidence that can be checked edge by edge against what its tools retrieved from the graph. On a knowledge graph built from the semiconductor-manufacturing literature, Prospector approaches the strongest off-the-shelf models given the same graph and tools, and it far exceeds off-the-shelf models answering closed-book or from a pre-selected portion of the graph. Although trained only on the semiconductor graph, Prospector transfers to unseen knowledge graphs from three other distinct fields: bio-inspired materials, per- and polyfluoroalkyl substances (PFAS) and materials properties. Served on a single GPU, Prospector answers faster and at a fraction of the cost of a hosted service, while keeping proprietary data on the owner's hardware. We find that the policy's latent state is indicative of its next move on both seen and unseen graphs, and modulating the latent state can steer the policy from one behavior mode to another. These results suggest a practical LLM + KG paradigm for scientific and engineering reasoning over private data, in which a compact model trained on a graph from published literature transfers its traversal to unseen graphs that stay with their owners.

WorkFactory: Turning Professional Intents into Executable Work for Scalable Agent Training

Ziyang Zeng, Tingyang Chen, Jia Deng, Qiyao Wang, Shuo Tang, Yike Yang — Wed, 28 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 73%, Agents 53%

Scaling supervised training for working agents requires diverse professional tasks that yield reliable expert demonstrations. However, missing or inconsistent task information can obstruct teacher execution and weaken the resulting supervision. We propose WorkFactory, a scalable framework for constructing coherent and executable professional tasks. Starting from professional intents, information-level planning builds an information support graph to guide task instantiation, explicitly representing the dependencies and locations of information required for the intended outputs. Probe executions then identify support failures and guide repairs while preserving the intended work. We construct 112,686 tasks spanning 612 occupations and use the collected teacher trajectories to fine-tune Qwen models at 4B, 27B, and 35B-A3B scales, yielding the WorkPilot family. All three models improve over their respective base models across four professional-work benchmarks, with WorkPilot-35B-A3B improving from 879 to 1565 Elo on GDPval-AA v2.1. Scaling analyses reveal model-dependent diminishing returns, with execution length first increasing and then decreasing. Controlled component analyses show that planning and repair reduce probe-diagnosed task environment defects and improve downstream performance. The data and models are available at https://github.com/IQuestLab/WorkFactory.

Clearing the FoG: Training KG Agents to Retrieve the Evidence Behind Their Answers

Zongyuan Sun, Shuai Zhao, Yingjun Du, Ke Shu, Ziyue Zhu, Haoran Luo — Wed, 28 Oct 2026 00:00:00 +0000

92% · New model 92%, Hallucination 70%, Evaluation 61%

Large language models (LLMs) can answer many factual questions correctly, but a correct answer is not necessarily a supported one. For an LLM agent, a knowledge graph (KG) plays two roles: as a source of answers, it supplies facts that the model lacks; as a source of evidence, it supplies the supporting triples that make an answer traceable and auditable. However, existing KG agents treat the KG only as a source of answers: their training data carries answers but not evidence, and their training rewards check the answer but not the retrieved evidence, so agents often answer correctly without retrieving the supporting triples. We propose FoG, a framework that trains KG agents to retrieve the evidence behind their answers. FoG consists of evidence-grounded synthesis, a KG interaction agent, and an evidence-coverage process reward (ECPR): the synthesis recursively hides explicit entities behind verifiable sub-questions, so that every question carries its supporting triples by construction; the agent records the triples returned by the KG; and ECPR credits each supporting triple to the turn that first retrieves it. Experiments on three KGQA benchmarks demonstrate that FoG enables 3B--4B open-source agents to outperform substantially larger baselines in answer accuracy and evidence coverage, with zero-shot cross-KG transfer.

Don't Pick a Winner: Knowledge Conflicts Need a Prior, Not a Rule

Minh Pham Van, Cong Tran — Wed, 28 Oct 2026 00:00:00 +0000

91% · Evaluation 91%, Hallucination 63%, New model 32%

A retrieval-augmented language model answers from two sources, the retrieved context and its own knowledge, and the two disagree whenever the context is incomplete or wrong. Existing methods settle the disagreement with a rule fixed at design time: follow the context, amplify it in proportion to how far it moves the model, or trust the model's memory. Each rule is confidently wrong in some deployment. In a decision-theoretic model of two sources, the score of any fixed rule is linear in the share of reliable contexts while the best attainable score is convex, so every fixed rule is beaten wherever it is not exactly optimal; and a symmetric divergence between the two sources, the signal adaptive decoding uses, carries no information about which one is right. Resolving a conflict therefore needs two inputs that no existing method keeps: evidence about each source separately, and a prior on how often the context is right. We test these claims on GRACE, a benchmark that intervenes on the knowledge-graph context of a fixed question, so that what the model knows, what the context says and what is true are measured separately. There, every baseline from four families is strong on right contexts or on wrong ones, never on both. We then build the two inputs: the model states its knowledge as notes, a small policy voices each note with a calibrated confidence, a reader answers in two voices, and a selector commits or abstains under a prior $\pi$ applied as an exact prior shift. With $\pi$ set to the true share of right contexts, the system is above the best baseline at every mixture of GRACE contexts, by up to 26 points. Its overall score is 45.3 and 35.0 (Qwen3.5-35B-A3B, Llama-3.1-8B), against 29.4 and 27.5 for the best baseline. On preregistered ClashEval it beats same-backbone CK-PLUG and TCR by 12 and 26 points. Where the model lacks the knowledge, the gains vanish, as the theory predicts.

WAKMan: Workload Aware KVCache Manager

Wed, 28 Oct 2026 00:00:00 +0000

91% · Efficiency 91%, Evaluation 32%

Prefix caching enables a Large Language Model (LLM) servers to reuse the Key-Value (KV) states and accelerate prefill computation stage for requests that share a prefix with earlier prompts, thus reducing latency and serving cost. With limited memory resource, the KV cache eviction policy becomes a decisive factor in determining the overall benefits of prefix caching. Production traffic may have single-turn, multi-turn, agentic, or a mixture of any combination of these request types. Our findings indicate that agentic steps return within seconds and favor recency, while popular prompts and long conversations return after many other requests and favor frequency. Across 20 mixes, no single policy is always best: vLLM's default strategy LRU provides 12% higher cache hits as compared to the frequency-oriented S3FIFO on agentic-heavy mixes but trails it by up to 22% elsewhere, and even a single workload parameter can flip the winner even for any particular mix. We present WAKMan, a workload-aware KV cache manager that picks its eviction policy from live traffic without knowing request types or reading prompts. It improves hit rate over LRU by 5 percentage points on average, with the lowest mean and worst-case optimality gap of all policies.

Correction or Manipulation? Measuring Epistemic Vigilance in LLM Agents Under User Pressure

Nardine Basta, Venus Haghighi, Dali Kaafar, Amin Beheshti — Wed, 28 Oct 2026 00:00:00 +0000

91% · Evaluation 91%

LLM agents that answer questions and act on behalf of users must decide when to yield to a user who disagrees. Disagreement may correct a genuine error or may be an attempt at manipulation. Prior work shows that models frequently abandon correct answers when challenged by users, yet it remains unclear whether they distinguish correction from manipulation or merely differ in their propensity to concede. We examine this question across seven LLMs from four vendors in two studies, one on the answers agents give and one on the actions they take as customer-support agents whose policy permits account changes only for verified users. Simulated users push toward correct and incorrect answers with graded pressure. Following research on epistemic vigilance, the pressure either appeals to the source, such as claimed authority, or offers content the agent cannot verify, namely a cited source. Drawing on signal detection theory, we measure on a set of benchmark questions how much more readily each agent concedes to correction than to manipulation, separately from its overall propensity to concede. Citations made five of seven models readier to concede and four worse at discriminating, while stronger wording added little. In the agentic study, five models never violated the verification policy, including models that readily abandoned correct answers, so concession on answers did not predict violations of policy. Ministral 3B violated the policy mainly under claimed authority, and Ministral 8B reported changes it had not made. An instruction to revise only on verifiable evidence chiefly shifted propensity to concede rather than improving discrimination. Concession rates alone are therefore misleading, and agents should be evaluated on both their discrimination and the actions they take.

NextEval: An Agent-Based Solver and Benchmark for Derivative-Free Optimization

Shuo Chen, Li Cui, Cunxin Huang, Ya-Feng Liu, Minghang Xie, Meng Xu — Wed, 28 Oct 2026 00:00:00 +0000

91% · Evaluation 91%, Agents 80%

Derivative-free optimization (DFO) addresses problems where derivatives are unavailable or costly to obtain. This setting often arises in simulation-based and experimental applications, where evaluating the black-box objective can itself be costly. Thus, it remains a central challenge in DFO to effectively exploit accumulated observations to select informative evaluation points under a limited budget. In this light, we introduce the NextEval solver, an agent-based DFO solver, which integrates the capabilities of a large language model (LLM) to guide the sequential selection of evaluation points. Specifically, we establish a system that combines an LLM agent with a numerical environment. The environment provides tools for analyzing past evaluations and executing numerical methods, while maintaining a shared evaluation history and method-specific state to support continuation across method switches. Using observations and tool feedback, the LLM agent selects tools, configures their parameters, and allocates the evaluation budget, while numerical methods compute and evaluate candidate points. Building on this system, we also introduce NextEval Bench, a benchmark protocol for comparing agent configurations on shared DFO tasks using the same numerical tools under matched evaluation budgets and information constraints. Numerical results on low-dimensional DFO problems demonstrate that the NextEval solver outperforms the state-of-the-art DFO baselines in function-evaluation efficiency.

Developing a Speech Language Model for Basque

Wed, 28 Oct 2026 00:00:00 +0000

91% · New model 91%, Evaluation 42%, Efficiency 31%

Although the recent development of LLMs has paved the way for multi-modal interaction, minority languages face difficulties keeping pace. This work presents the first open Speech Language Model for Basque, designed for low-latency, end-to-end spoken interaction. Based on the LLaMA-Omni architecture, the system integrates a Latxa model as its backbone, transferring its Basque capabilities to the spoken modality by fine-tuning several modules. To address data scarcity, the English VoiceAssistant-400K corpus was translated into Basque, and the audio was synthesized using a Basque TTS. Evaluation results confirm the system's viability, achieving 3.86/5 in content quality, a 9.94% WER for spoken intelligibility, and a 3.07 UTMOS for audio quality. This work demonstrates that advanced spoken interaction systems can be developed using modular architectures and synthetic data even in low-resource languages.

Do We Need an LLM to Know Why Sources Conflict? Decision Models for Knowledge-Conflict Typing in RAG

Parv Khatri, Harshul Jain, Mehak Kapoor, Tanya Sah — Wed, 28 Oct 2026 00:00:00 +0000

91% · Efficiency 91%, Evaluation 79%

When the sources retrieved for a query disagree, the right response depends on why they disagree: a question with several valid answers, a contested topic, stale information and a false source each call for different behaviour. Cattan et al. (2025) showed that telling a generator the conflict type improves its responses by about 24 points, but that predicting the type is hard even for frontier LLMs. We study whether typed decision models, which return probabilities over fixed options in a single forward pass instead of generating text, can do this classification fast and cheaply. Using the CONFLICTS benchmark and the original paper's prompt verbatim, we evaluate all 11 decision models served by one commercial inference gateway, from 7 providers, against Gemini models on identical inputs. Asked directly, the decision models reach 50.0--67.0% accuracy; 10 of the 11 match or exceed Gemini 2.5 Flash, the paper's model (61.3%), at 4--39x lower billed cost (2 were free) and 2.6--6.0x lower latency. The current Gemini 3.8 Flash (70.0%) is significantly more accurate than 8 of them, but sending each decision model's least confident quarter of queries to it gives a cascade statistically tied with it for 8 of the 11, at 25--37% of its cost. End to end, answers generated with the types of 10 of the 11 follow the expected behaviour as often as with the LLM's types. Decomposing the taxonomy into yes/no questions lowers accuracy for every model, and a repair fitted on one model does not transfer. Pre-specified diagnostics show that dates play no measurable role and that sensitivity to source order varies widely between models. Decision models detect a planted false source less reliably than the LLM as hard labels, but their probabilities still separate planted from clean sources. We release code, prompts and all model outputs at https://github.com/Parv17k/decision-models-conflict-typing.

DocXEval: A Systematic Evaluation of Explainability in Multimodal Document Understanding

Saifullah Saifullah, David Dembinsky, Stefan Agne, Andreas Dengel, Sheraz Ahmed — Wed, 28 Oct 2026 00:00:00 +0000

91% · Evaluation 91%

Multimodal transformer models have driven unprecedented performance breakthroughs in modern document understanding. However, the inherent black-box nature of these models remains a major obstacle to their safe deployment in high-stakes sectors. Despite significant progress in explainability for vision and language tasks, explainable AI (XAI) research tailored to multimodal document understanding has received little attention. To address this gap, we present \ourframework{}, the first comprehensive explainability framework that adapts existing XAI methods and evaluation metrics to the multimodal (text, layout, and image) document understanding context, while introducing efficient multi-target adaptations of both methods and metrics for domain-specific tasks such as semantic entity recognition. Using \ourframework{}, we systematically evaluate 13 XAI methods across 3 document understanding tasks, 8 datasets, 4 models, and 17 evaluation metrics, and find that only a few methods perform consistently across all evaluation criteria, with Occlusion emerging as the most faithful and interpretable method overall. Through a dedicated meta-evaluation, we observe moderate to strong agreement among several faithfulness and complexity metrics while identifying certain ones that produce unreliable results. Finally, through multiple dataset-wide diagnostic analyses, we uncover key insights into model behavior, including the contribution of individual modalities across tasks and labels, spatial attribution patterns around target entities, and potential dataset biases. Overall, this work serves as both a benchmark and a practical guide for future XAI research in multimodal document understanding. To support reproducibility, we publicly release our code at https://github.com/saifullah3396/docxeval/.

CaLLM: Continual Adaptation of LLMs via Non-Parametric Adapter Routing

Bilge Celik, Anna Vettoruzzo, Joaquin Vanschoren — Wed, 28 Oct 2026 00:00:00 +0000

91% · New model 91%, Efficiency 34%, Evaluation 30%

Continual adaptation of large language models (LLMs) to evolving task distributions while mitigating catastrophic forgetting remains an open problem. Existing strategies often struggle to scale or lack a robust mechanism for zero-shot task-to-expert assignment. We introduce CaLLM, a framework that resolves these tensions by combining metric-based meta-learning with modular parameter isolation via lightweight adapters. First, CaLLM projects each input batch into a task-agnostic embedding space to find its nearest prototype. It then routes the batch to the specific adapter specialized for that data distribution. This formulation provides (i)~controllable cross-task interference by design, (ii)~constant computational complexity regardless of task count, (iii)~seamless integration of novel distributions through prototype expansion, and (iv)~regime agnosticism as the local routing decision enables a unified approach to both online and offline settings. Our experiments demonstrate that this routing mechanism discovers semantically coherent task groupings while facilitating constructive interference between similar tasks, and keeping overhead at a minimum. Across diverse and fine-grained continual learning benchmarks, CaLLM consistently outperforms traditional isolation and replay methods, establishing prototype-based modular routing as a principled and scalable approach to lifelong adaptation of LLMs.

RPSD: Retrieval-Privileged On-Policy Self-Distillation with Optimal Transport

Dongyue Li, Zebang Cheng, Zhongyi Deng, Hongboxu, Yi Bin, Fei Ma — Wed, 28 Oct 2026 00:00:00 +0000

91% · New model 91%

Sequential recommendation predicts a user's next interaction from historical behavior. Recent LLM-based methods recast this task as generative recommendation and increasingly adopt reinforcement learning such as GRPO for post-training. However, these methods suffer from two limitations, namely credit-blind optimization, where a sequence-level reward is shared across all generated tokens, and collaborative-blind supervision, where each training signal is derived from a single user and ignores behavioral patterns shared across similar users. To address both issues, we propose Retrieval-Privileged on-policy Self-Distillation (RPSD). For credit-blind optimization, we introduce on-policy self-distillation (OPSD), in which a teacher with privileged knowledge re-evaluates the student's on-policy response, and the token-level confidence difference between the student and teacher reshapes the shared sequence-level advantage into token-specific credit. However, the privileged knowledge in vanilla OPSD is still derived from a single user, leaving collaborative-blind supervision unresolved. We therefore ground the teacher in collaborative behavioral evidence from similar users. Specifically, we use temporally aware optimal transport (OT) to compare interaction trajectories across users by jointly considering item similarity and temporal structure, and further apply diversity-aware selection to retain complementary trajectories as privileged evidence. In this way, on-policy self-distillation provides fine-grained token-level credit, while OT enriches the teacher with structured cross-user evidence. To the best of our knowledge, this is the first work to bring on-policy self-distillation to sequential recommendation and to use cross-user behavior as its privileged signal. Extensive experiments across multiple real-world datasets demonstrate that RPSD achieves state-of-the-art performance.

Uncovering Why LLMs Fail: From Internal Mechanisms to Adversarial Reasoning Trajectories

Wed, 28 Oct 2026 00:00:00 +0000

91% · Evaluation 91%, Hallucination 37%

Large language models (LLMs) are usually evaluated by final-output correctness or safety, yet an endpoint label does not reveal how a failure develops throughout the reasoning process. We therefore ask whether ordinary reasoning failures and adversarially induced failures can be characterized using the same internal coordinates. Towards this end, we in this paper study a white-box diagnostic framework CauTrace to disclose the root-cause of why LLM fails from three fundamental inspections: Functional States, Propagation of Representation, Temporal-State Coordination and Reasoning Trajectory. First, seven functional states are creatively defined to represent the support and constraint processing in LLMs. Second, the propagation of representation change measures how the changes among the seven states persist across subsequent Transformer blocks. Third, temporal-state coordination and trajectory topology characterize when the commitment and verification occur relative to the support, and how the dominant state sequence evolves over the full inference process. Finally, we evaluate CauTrace on both clean reasoning failures and paired adversarial perturbations. Across Qwen and LLaMA-family models, the failed traces show weaker Attention-MLP agreement on support-state updates and less verification activity relative to commitment, while the exact representation-level signatures and timing shifts vary by architecture. Held-out repair and steering tests demonstrate consistent operational signals that were in line with our claims for certain models and specific settings, with aligned steering producing the predicted directional changes in process measures. Overall, CauTrace provides a unified diagnostic framework for analyzing failure dynamics through changes in support, commitment, and verification, which would shed light on the in-depth research on LLM security.

Same Verdict, Different Readers: Personalizing Fact-Checking Explanations from Relative Preferences

Haokun Wen, Xuemeng Song, Mong-Li Lee, Wynne Hsu — Wed, 28 Oct 2026 00:00:00 +0000

90% · New model 90%, Hallucination 58%, Evaluation 53%

Fact-checking explanations help Web users understand why a claim receives its verdict, yet the same explanation may not satisfy readers with different backgrounds and preferences. We formulate Faithful Personalized Fact-Checking Explanation Rewriting, which adapts an existing explanation to individual readers while preserving factual fidelity. Rather than relying on reader-written reference explanations, we learn personalized rewriting from the relative preferences of LLM-simulated readers. However, this presents two challenges: distinguishing persona-specific preferences from general explanation quality and learning beyond a limited set of predefined explanation styles. To address these challenges, we propose Personalized Explanation Rewriting via Preference Learning (PERPL), a framework consisting of two components. Simulated Preference Construction derives persona-specific supervision from simulated reader preferences through cross-persona reciprocal pairing. Progressive Preference Learning first trains a lightweight rewriter on the constructed supervision and then refines it through exploration-guided preference optimization, enabling it to discover preferred explanations beyond the predefined styles. Extensive experiments demonstrate that PERPL outperforms much larger general-purpose LLMs in personalized preference alignment while maintaining high factual fidelity.

GEPA-Curator: Curated Learning for Self Improving Agents

Zack Yu, Nicholas Lee, Minjae Lee, Lakshya Agarwal, Luca Manolache, Wonjun Kang — Wed, 28 Oct 2026 00:00:00 +0000

90% · Agents 90%, New model 70%, Evaluation 43%

Large Language Models (LLMs) are increasingly used in multi-turn agentic workloads that require sustained interaction with complex environments. Existing reflective prompt optimizers such as GEPA learn from failed rollouts, but agentic failures can also come from prompt weaknesses, harness limitations, stochastic execution, or noisy judges. We argue that such failures may not provide useful learning signal, and may even be harmful; and thus it is essential to curate which tasks are provided to the prompt optimizer for better performance. We introduce GEPA-Curator, a framework that augments the prompt optimizer to select which agent experiences should shape prompt optimization. In particular, GEPA-Curator consists of two components: task curation, which filters out tasks that are likely to provide noisy or unreliable learning signals, and prior success retrieval, which selectively retrieves past successful trajectories for a task and provides them as additional context to the prompt optimizer. We evaluate GEPA-Curator on GoBrowse, ALFWorld, AppWorld, and $\tau$-bench, using both Gemini 3.1 Pro and Claude Opus 5 as optimizer models. GEPA-Curator matches or exceeds the mean performance of both baselines, with gains of up to 20.7% over GEPA and 21.2% over ACE.

★ BenchBench-Cloning: Evaluating Molecular Cloning Design and Protocol Development

Aditya Sivakumar, Peter Ciaccia, Hassaan Qadir, Ian Hsu, Ashu Singhal — Tue, 27 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Molecular cloning is one of the most routine wet-lab workflows, yet existing evaluations of language models in biology mostly test knowledge recall or isolated steps rather than end-to-end experimental planning. We introduce BenchBench-Cloning, a benchmark for large language models comprising 65 molecular cloning tasks grounded in sequence records, laboratory notebooks, and published experimental studies. Planning a molecular cloning experiment requires defining a set of constructs needed for a scientific objective and a construction route to produce them. BenchBench-Cloning contains 50 protocol-development tasks, in which agents produce an executable protocol for a specified construct, and 15 construct-design tasks, in which agents determine the constructs a scientific objective calls for. Protocol-development tasks are evaluated by execution in a cloning simulator; construct-design tasks are evaluated with task-specific deterministic checks and, where needed, rubric-based LLM-as-a-judge evaluation. Across eight systems with repeated attempts per task, GPT-6 Astra scores highest at 30.6%, with the others between 8.6% and 22.2%, and the benchmark remains unsaturated under best-of-ten sampling. As integrating individual cloning decisions into complete, valid plans remains challenging even for frontier models, we expect BenchBench-Cloning to serve as a useful benchmark for tracking progress in experimental planning.

★ The LALA Benchmark: Evaluating Literary and Linguistic Analysis in LLMs

Tue, 27 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

We introduce the Literature and Linguistics Analysis Benchmark (LALA), comprising 236 open-ended questions written by experts in linguistics and literary studies, together with expert-provided ideal answers. By comparing models within families and at similar training scales using human and LLM judges, LALA identifies differences in linguistic competence associated with post-training, offering a framework for evaluating future models and guiding targeted improvements. We administered questions to 17 language models and evaluated their responses using both human expert assessment and two LLM evaluators, Sol and Qwen. Base models displayed little of the linguistic competence assessed by the benchmark, whereas their post-trained versions performed markedly better, suggesting that post-training plays a central role in making linguistic knowledge accessible and usable. Parameter count, by contrast, showed no statistically significant association with partial-credit performance, and competencies remained unevenly distributed across linguistic domains. We interpret this pattern as evidence that post-training adds linguistic competencies piecemeal, without producing a general linguistic intelligence capable of extending them beyond the learned policies that support them. Models performed best on logical relations between statements and selected discourse-level tasks, while even the strongest post-trained models continued to struggle with ambiguity, literary style, reference, and misleading assumptions. Human and Sol evaluations produced nearly identical model rankings in both the original and ReliableEval runs, supporting the replicability of the comparative results.

★ Task Success Is Not Enough: Side-effect-Aware Evaluation of Tool-Using Language Model Agents

Jiaju Huang, Shaobin Chen, Xinglong Liang, Xinyu Ma, Hao Yang, Wei Ke — Tue, 27 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 98%

Tool-using language model agents are usually evaluated by whether they finish the user's task. That metric misses a common failure: an agent can complete the requested work while also taking an unauthorized side-effecting action, such as writing to the wrong object, expanding the scope of the request, or contacting an unintended recipient. We call this failure mode collateral damage (CD). We introduce SABench (Side-effect-Aware Benchmark), a benchmark of 201 email, calendar, and ticket tasks. Each task crosses one of four state structures (independent, dependent, externally reachable, cross-tool) with one of four trap families (ambiguity, mis-targeting, cascade harm, overreach), and each task has a deterministic trace-based oracle for both goal satisfaction and authorization compliance. Across 2,211 model-task pairs from 11 frontier LLMs spanning major commercial and open model families, 15.2% of all runs incur CD, with risk concentrated in ambiguity tasks (38.8% violation rate) while non-ambiguous traps average 8.4% CD. Among goal-satisfied runs, 12.8% still produce unauthorized writes. Cross-tool and dependent states have about twice the CD rate of independent states. CD rates vary by 4.7x across models at comparable goal rates, which suggests that task success alone is a poor proxy for safe tool use. The SABench task corpus is released at https://huggingface.co/datasets/Phaedrus/sabench; the traces, oracle, code, and analysis pipeline are released at https://github.com/hjj194/sabench-artifact.

★ R2A-Sales: A Policy-Atom-Matched Benchmark for Rule-to-Action Gaps in Goal-Directed Sales Agents

Tue, 27 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 92%, Hallucination 51%

LLM-based sales agents are often evaluated on static policy questions or task success, but deployment asks a sharper question: when a model can state a rule, will it choose compliant, tool-grounded actions under commercial pressure? We introduce R2A-Sales (Rule-to-Action), a policy-atom-matched benchmark for this question. Each item pairs a static probe with a multi-turn sales scenario sharing the same versioned policy atom, facts, evidence, behavioral constraints, and tool contract, so differences cannot be attributed to policy-content mismatch. In the interactive arm, a code-owned ledger verifies committed tool effects and terminal states; unsupported action claims receive no credit. We report static compliance $K$, grounded action compliance $A=\Pr(\text{safe terminal}\land\text{no hard violation}\land\text{non-echo})$, and the matched gap $G=K-A$. The frozen release spans three controlled Chinese-language sales domains, 24 policy atoms, 42 semantic cores, and 2,016 matched pairs. The confirmatory study audits 21,888 runs over 19 cores, covering eight backbones plus four Qwen3-8B prompt treatments (12 configurations). Static compliance is near ceiling $(K=89.9\%\text{--}99.1\%)$, yet eight-backbone action compliance is only $A=3.0\%\text{--}17.3\%$, giving $G=80.4\text{--}92.5$ percentage points. Since $A$ also requires safe termination, $G$ upper-bounds rule-to-action degradation rather than isolating pressure alone. No prompt-only treatment dominates all metrics. R2A-Sales shows that sales-agent reliability must be measured at committed action, not rule recital or task completion alone.

★ InferPriv: Evaluating Inferential Privacy Leakage in LLM Assistants

Pengrun Huang, Yiwei Fu, Luoxi Meng, Kamalika Chaudhuri — Tue, 27 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Large language model (LLM)-based personal agents are increasingly deployed to handle everyday tasks on users' behalf. These agents have access to private user information, while their tasks involve communicating with third parties; this raises a critical privacy concern that agents may inappropriately reveal sensitive user information to these recipients. Existing benchmarks primarily assess direct privacy leakage, where an agent explicitly states a pre-specified secret or attribute. However, they do not capture \textit{inferential leakage}, where the model leaves indirect cues in the reply that implicitly leaks sensitive attributes. We introduce \textsc{InferPriv}, a benchmark that measures whether an honest-but-curious recipient can infer a user's sensitive attributes from an LLM's reply on routine task. Evaluating five frontier models, we find that inferential leakage substantially exceeds direct leakage across models and prompts. Under a privacy-aware system prompt, GPT-6.1-Sol, Sonnet-5.5, and Gemini-3.8-Flash achieve near-zero direct leakage, yet the inferential leakage rate remains between 16.2--25.2\%. These findings suggest that evaluating explicit leakage alone can substantially understate the privacy risks.

★ Omni-InteractScope: Benchmarking Proactive and Agentic Interaction in Continuous Omnimodal Streams

Yinsong Yan, Ling Wang, Yuxuan Wang, Yifan Yang, Ruiyang Xu, Junming Lin — Tue, 27 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 61%

Omnimodal large language models (Omni-LLMs) are increasingly expected to assist users while observing continuous audiovisual streams. However, existing evaluations provide limited insight into how reliably these models connect stream evidence and user needs to timely, sustained interventions. We introduce \textbf{Omni-InteractScope}, a benchmark of 1,479 human-annotated, cross-reviewed interactions with 1,790 temporal targets across eight task groups. It links input modalities, event triggers, intervention types, and user needs, covering event-timed tool invocation and query-free assistance alongside notifications, warnings, state tracking, and continuous responses. Our experimental results on the proposed benchmark reveal three critical limitations of current Omni-LLMs: (1) \textbf{Timely tool invocation remains unreliable}: all 12 evaluated models produce an on-time first response for at most 40\% of tool-invocation targets, and joint timing-and-tool accuracy reaches at most 32.0\%. (2) \textbf{Query-free reminders remain poorly covered}: all evaluated models achieve lower timely reference-target recall on query-free reminders than on explicit questions, measured on their respective target sets. Reminder recall never exceeds 23.0\%. (3) \textbf{End-to-end guidance coverage remains limited}: nine models cover none of the 26 guidance sequences in full, and the remaining three cover only one to three sequences each. Targeted controls further show that tool-invocation scores alone do not establish event grounding and that reminder coverage depends on response history and frequency. These findings motivate developing and evaluating streaming assistants for timely action, initiative under general user needs, and coverage of entire response sequences.

★ Patient-specific rubric-based evaluation of large language models across longitudinal cancer care

Wei-Wei Zhang, Pengkai Wang, Yan Li, Zhijie Sang, Guan-Qun Zhou, Jun-Yan Li — Tue, 27 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Hallucination 55%

Large language models (LLMs) show promise in specialty medicine, yet existing evaluations largely focus on isolated clinical encounters rather than tracking how clinical responsibilities evolve across a longitudinal care pathway. Here we introduce NPC-RubricBench, a phase-aware benchmark for evaluating multistage clinical reasoning across the cancer care continuum, using nasopharyngeal carcinoma as a defined disease context. Curated from longitudinal records of 405 patients across two cancer centers, the benchmark comprises 941 patient-phase units across six sequential care phases and 17,688 patient-specific atomic criteria spanning five evaluation dimensions, with each assessment strictly restricted to decision-time clinical evidence. Evaluating 30 general-purpose and medical LLMs revealed that strong diagnostic performance does not carry over to downstream care. We evaluated 30 general-purpose and medical LLMs, showing that high diagnostic performance did not translate into improved downstream care. Instead, the dominant failure modes reversed across phases. Explicit decision errors dominated anatomical staging at 48.2%, whereas decision omissions dominated initial work-up and post-treatment follow-up at 33.4% and 30.4%, respectively. In the evaluated model set, medically adapted models showed higher omission rates than general-purpose models. Moreover, varying query granularity exposes a critical specification trade-off: broad open-ended requests degrade Decision Quality across all 30 models, while more detailed, checklist-like prompts raise Decision Quality but reduce Evidence Faithfulness and Safety Adherence. Furthermore, NPC-RubricBench demonstrates that static diagnostic benchmarks are insufficient to ensure consistent clinical reliability over time, highlighting the importance of phase-adaptive safeguards in multistage medical AI.

★ A Benchmark for Controllable Speaking-Style Captioning in Audio-Language Models

Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 96%

Speaking-style captioning (SSC) aims to generate natural language descriptions of how speech is delivered, capturing paralinguistic attributes such as vocal timbre, prosody, and expressivity. Many downstream applications, including conversational AI agents, controllable speech generation, and large-scale audio annotation, require dimension-specific style captions that describe targeted aspects of speech (e.g., speaker traits, emotion, or delivery style), rather than a single undifferentiated description. However, existing work lacks a unified formulation for controllable speaking-style captioning. We address this gap by formulating SSC as an instruction-following audio-language task, where explicit instructions specify the speaking-style dimensions to be described. We introduce StyleInstructCaps, the first standardized benchmark for controllable speaking-style captioning in audio-language models. StyleInstructCaps provides a dataset and evaluation framework that measures metadata groundedness, hallucination, instruction-following ability, generalization to unseen instructions and audio datasets, and speaker-style consistency. Across nine state-of-the-art audio-language models, we find that strong instruction-following ability does not necessarily imply grounded speaking-style understanding, and that SSC-specific instruction-diverse training substantially improves groundedness, controllability, and speaker consistency.

★ AOS-Bench: Can LLM Agents Manage Dynamic Workloads?

Erle Zhu, Genrui Zhang, Keyang Hu, Mingyu Su, Hongning Wang — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 48%

Coding agents are moving from handling one request at a time to serving as persistent assistants to which users hand a stream of work: a build, a data-processing job and a bug fix may all be pending on the same machine at once. Such an agent must decide when to take on new tasks, how to delegate and overlap them, and how to stay within the machine's CPU and memory. Existing benchmarks cannot measure this, because they evaluate each task in isolation. We introduce \benchname{}, which evaluates an agent system as a workload manager: verified tasks arrive over time and must be delivered under shared CPU and memory limits, and the same tasks run one at a time serve as a per-system reference. Across five model--harness pairs, isolated task success does not carry over evenly. On heavy workloads every system loses 13 to 23 points. On medium workloads, systems span 11 points in isolated success but 26 points in stream success, and two systems 0.6 points apart in isolation deliver 54\% and 68\% of the same tasks. Execution-policy interventions trace much of this gap to task intake: serial dispatch leaves 73\% of heavy-workload tasks unclaimed. Yet more elaborate management does not necessarily help. A fixed concurrency of two comes within four points of the best agent-managed policy on both medium and heavy workloads, and under every policy median CPU use stays below 30\% of the provisioned cores while tasks wait. These results identify workload management as a distinct capability for agents, and \benchname{} as a way to measure it.

★ Position, Role or Label? A Controlled Study of Instruction Conflicts in LLM Agents

Anish Laddha, Nitesh Pradhan, Gaurav Srivastava — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 64%

LLM agents take instructions from the operator, repository configuration, user, tool results and saved memory, which often conflict. Instruction-hierarchy benchmarks ask whether the higher-ranked source wins, but a source sets position, role and label at once. First, we build WhoWins, 1,080 code-scored scenarios in five task types, each with up to five mutually exclusive instructions, and seven controlled experiments that vary only how instructions are delivered. Second, we study 14 open-weight models (0.6B to 122B, thinking off and on) over 1,887,356 scored replies, checking main findings on held-out scenarios. Inside one message the first of two conflicting lines wins 72.42 to 94.66%, and across messages it keeps 22.95 to 62.65%. A policy sent first in the system (or developer) message loses most conflicts with later lines in every Qwen model but wins most on gpt-oss when it acts. In tool and file-write tasks it wins less often than the same text sent first as a labelled user message, in every model-mode. Third, we test whether pairs forecast larger conflicts; forecasts improve in every model-mode once they know where the task line is. Finally, by our keyword patterns at most 2.92% of decided pair replies flag or hedge a conflict.

★ NewsBench: Expert-Grounded Evaluation of Epistemic Quality in AI News Reporting

Matt Wilde, Jillian Fisher, Andrew B. Hall, Kathryn Salam, Robbie Goldfarb — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 58%

As frontier AI systems increasingly serve as sources of political and current-events information, evaluating the epistemic quality of their responses is now especially consequential. Existing evaluations often focus narrowly on ideological alignment, rely on measures not validated for this setting, or depend on reference labels derived from public opinion or non-expert annotation, potentially producing misleading or miscalibrated results. To address these limitations, we introduce NewsBench, an expert-grounded benchmark that reframes evaluation around editorial standards developed by senior practitioners in journalism, policy, and intelligence analysis. NewsBench operationalizes quality across three dimensions, source quality, factuality, and neutrality, using a two-phase pipeline in which domain experts first define and calibrate evaluation guidelines, which are then transferred to automated AI judges validated against expert-adjudicated gold labels. The benchmark includes approximately $3,000$ evaluation prompts, with roughly $500$ refreshed monthly to capture emerging current events, roughly $2,500$ human expert-annotated items for calibration and validation, and a suite of AI judges designed to enable scalable evaluation. To validate our methodology, we report both inter-rater reliability among human experts and AI judge--human agreement across all three dimensions. Finally, we apply NewsBench to four frontier models, revealing distinct and non-overlapping failure profiles in sourcing, factuality, and neutrality.

★ AgentOrch: How Close Can Global Orchestration Come to Step-by-Step Interaction?

Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 89%, Efficiency 81%

Enterprise workflow agents typically rely on step-by-step interaction (Step-by-Step): the agent executes an action, observes its result, and invokes a large language model (LLM) again to decide the next step. Global orchestration (Global) instead generates a complete executable workflow in advance and executes it without further LLM calls, substantially reducing inference overhead, but it remains less accurate than Step-by-Step. We study how close Global can come to Step-by-Step and what narrows the remaining gap. We introduce AgentOrch, a benchmark that directly compares global orchestration with step-by-step interaction on 650 enterprise SOPs and 3,220 runtime variants under matched tools, environments, and evaluation. Step-by-Step remains more accurate for every evaluated model, but the gap has shrunk by about 79% from early models to the strongest recent model. We identify three routes to further narrowing this gap. Incremental workflow construction improves Global accuracy by an average of 9.98 percentage points (pp) without runtime execution feedback. Revising complete workflows with execution feedback improves all evaluated models, including 6.13 pp on held-out runtime variants. Targeted training on complete workflows improves accuracy by an average of 12.18 pp across all five trained models. Although a gap remains, global orchestration offers a promising path toward enterprise agents that deliver higher accuracy with fewer LLM calls.

★ BoardSafety: AI-Ready Data from Manufacturer Documentation for Development-Board Safety

Linjian Yang, Xinyan Wang, Kunpeng Liu — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 32%

The development of large language models (LLMs) are capable to assist with hardware experiments. In this scenario, this raises a practical question: Can LLMs determine whether a proposed operation is safe for a specific development board? Documented operating limits and safety requirements are board-specific, even across products from the same manufacturer. This variation makes board-specific safety judgments by an LLM difficult to evaluate. To address this problem, we introduce the BoardSafety Benchmark. The benchmark is built from vendor datasheets and hardware guides. For data transformation, these documents are converted into a board-specific safety rule base and are then used to create labeled operating scenarios. Each data item pairs a board and a proposed operation to an applicable safety rule with ground-truth label, indicating the level of safety. The categorized level for each operation includes safe, unsafe, or insufficient-info. In addition, the benchmark covers diverse operations, board-specific cases, boundary conditions, and missing-information cases. We evaluate a series of models by varying their access to board-specific rules and the use of structured-rationale prompting. Our results demonstrate that providing the safety rules generally improves accuracy, whereas structured-rationale prompting produces gains that vary across models and rule-access conditions. Overall, the evaluated models cannot yet reliably determine whether an operation is safe, which implies that critical exploration is required for this field.

★ Small Edits, Shifting Decisions: Benchmarking LLM Accuracy and Reliability on French Input Variants

Jiaming Yu, Xiaotian Xie, Jian Ma, Yingjie Li — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

French users may omit apostrophes and diacritics or make keyboard typos in everyday input. The effects of these common variations on language model performance and prediction reliability remain insufficiently understood. To address this gap, we construct a controlled French surface-variant benchmark covering seven task types, with 4,200 original-low-high triplets (12,600 input instances). Each high-intensity variant retains all low-intensity edits and adds further changes. We evaluate six models from two LLM families. The results show that accuracy decreases and Brier score and negative log-likelihood (NLL) increase with perturbation intensity. Larger models show smaller accuracy losses under high intensity, but increases in Brier score and NLL do not consistently decrease with scale. Keyboard typos cause the largest losses in accuracy and the quality of predicted probabilities at the predefined high-intensity settings. Finally, calibration temperatures fitted on original inputs transfer effectively to surface variants. Combining calibration with input restoration further improves Brier score and NLL over either method alone while preserving restored accuracy. Our benchmark provides a unified framework for jointly evaluating task accuracy and prediction reliability under common French surface variations.

★ AlphaIdeaBench: Benchmarking Automated Research Idea Generation under Feasibility Constraints with Peer Review Alignment

Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 37%

Research ideas are always proposed under resource constraints: a plan must fit the available compute, API credit, and time. Yet evaluations of research-ideation agents mostly score unconstrained novelty or overlap with a published paper, and their LLM judges are rarely checked against peer review. We introduce AlphaIdeaBench, a benchmark of 100 ideation tasks (50 workshop calls for papers and 50 open problems stated in surveys), each paired with an explicit resource budget and a shared proposal schema. An LLM judge scores each proposal on 12 dimensions covering validity, novelty, and feasibility. To validate the judge, we compare its scores on four novelty axes (problem, method, scenario, and insight) with scores derived from reviewer comments on 600 papers from ICLR 2026, NeurIPS 2025, and ICML 2025. For problem and method novelty, Gemini 3.7 Flash is within one point of the reviewer-derived score on 94-97% of papers; agreement on scenario novelty is relatively lower. We evaluate eight published harnesses on a shared backbone and, within one harness, eight backbone LLMs. CoI-Agent obtains the highest overall score, narrowly ahead of SciMON, which leads on method and insight novelty. All harnesses score low on scenario novelty and rarely change the problem setting. The backbone largely determines which open problem a proposal targets, whereas the harness determines whether a proposal addresses one problem or combines several. Weaker backbones such as Llama 4 Scout fail to carry out that harness's strategy and score low on budget feasibility. Finally, when only the budget is varied on three topics, a larger budget changes the experimental setting on compute-intensive topics but only adds tests on an inexpensive one. We release the tasks, rubrics, and scores to foster standardized, reproducible benchmarking of future research-ideation agents under constraints.

★ InfiniteWorld: Separating Visual Evidence from Familiar Physics

Tong Zhang, Yuxuan Cheng, Yun Peng — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 30%

Physical reasoning requires adapting familiar expectations to the dynamics of the observed world. Existing benchmarks establish broad physical capabilities, yet an overall score leaves unclear whether answers follow visual evidence when familiar rules no longer apply. We introduce InfiniteWorld, a benchmark that pairs physical-reasoning questions with simulator videos whose physical parameters are systematically altered. Its generation pipeline connects simulation, scene screening, paired video translation, question construction, and input checks. Five tasks assess law identification, parameter estimation, temporal reasoning, analytic counterfactual comparison, and law induction. The evaluation defines five matched input settings for each question, comprising text-only evaluation and ordered or shuffled observations from simulator and translated videos. We evaluate a diverse set of video-language models under these matched settings. The results expose a persistent preference for familiar physical rules. Frames improve performance for most models, but gains differ across tasks. Strong accuracy on Earth-normal questions coexists with poor recognition of altered dynamics, and Earth-normal answers remain common when the observed physics changes. Task scores and semantic agreement further distinguish matching physical factors from matching their amount, and correct answers from repeated errors. InfiniteWorld makes these behaviors measurable through paired changes in the evidence supplied for the same question, separating visual evidence from familiar physics.

★ IntentDelta: A Human-Constructed Benchmark for Evolving User Intent

Jieyu Zhou, Zijian Wang, Yun Wang — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 89%

Language-model agents must adapt as users revise their requirements and priorities during complex tasks. We introduce IntentDelta, a human-constructed multi-turn benchmark with 638 trajectories and 3,766 turns across WebShop and TravelPlanner. Each turn pairs a user utterance with a cumulative intention state and human reference actions. The trajectories are cross-verified and cover feasibility- and preference-driven changes in constraints and priorities. We evaluate eleven agents on intention recovery and action quality. Average Intention Recovery is 33.1\% in TravelPlanner and 44.7\% in WebShop, while average Action Score is 31.6% and 41.0%, respectively. Agents struggle to preserve earlier requirements and distinguish soft preferences from must-have constraints. Correct intention recovery also does not guarantee successful execution, highlighting challenges in translating evolving user intentions into actions.

★ Continual Learning Exam: Benchmarking Long-Horizon RSI and Experiential Learning in LLM Agents

Haoyu Zhao, Sitian Tu, Yuxuan Huang, Huichi Zhou, Zhangyue Yin, Jiahao Huang — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Continual learning is the ability of LLM agents to improve over time by acquiring, refining, and applying knowledge from experience while retaining useful prior knowledge. It is essential for agents learning over long time horizons, yet current evaluations largely assess individual task performance rather than improvement through accumulated experience. We introduce Continual Learning Exam (CLE), a large-scale benchmark developed by 50+ domain experts to measure experience-driven improvement in realistic professional workflows. CLE comprises 120 human-authored tasks, each organised as a five-level ladder, yielding 600 levels spanning 89 professional roles across 12 occupational domains. Agents must pass each level to unlock the next, with later levels building on experience accumulated during earlier assignments. An evaluation agent assesses each submission against human-authored rubrics. We compare a single-submission setting with one that provides diagnostic feedback after failure and allows one revision. Across 17 harness-model configurations using frontier LLMs, feedback and revision yield a median completion gain of 5.3 percentage points. Yet the best-performing system clears only 47.7% of levels and completes all five levels in just 23.3% of tasks, highlighting the difficulty of sustained learning from experience for existing frontier LLMs. As such, the dataset can also be used to post-train LLMs to strengthen their continual learning capabilities.

★ Can LLMs Trust Appropriately? Evaluating Calibrated Reliance on Imperfect Partners

Mengdi Chu, Jintao Huang — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Large language models increasingly collaborate with humans and peer AI agents whose contributions may be incomplete, inconsistent, or incorrect. Effective collaboration requires calibrated trust, whereby decisions to rely on, verify, or assist a partner adapt rationally as empirical evidence accumulates. We investigate whether models translate partner evidence into appropriate reliance policies. We introduce a benchmark of 324 scenarios across 23 collaboration categories. Each scenario involves three sequential decisions with explicit transition dynamics, feedback emissions, and cost schedules, permitting exact analytical evaluation of choices against their expected consequences over the remaining interaction. Evaluating seven models alongside 36 controlled-history probes, we dissociate evidence sensitivity from normatively appropriate action. While the strongest frontier model approaches near-optimal decisions, other models exhibit recurring pathologies: procuring excessive safeguards, purchasing verification tests with zero expected value of information, prioritizing immediate savings over cooperative learning, and misallocating assistance to suboptimal channels. Supplying explicit posterior distributions resolves certain errors while introducing new failure modes. Calibrated trust in autonomous agents critically demands coupling empirical evidence with forward-looking decision utility.

★ $\Delta\tau$-Bench: Pushing Agents Toward a Moving Capability

Shuofei Qiao, Xuanguang Pan, Yijun Chen, Wendi Feng, Bin Wu, Runnan Fang — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 95%

In this paper, we introduce $\Delta\tau$-Bench, an executable agent benchmark that models the structural evolution of large tool ecosystems. In $\Delta\tau$-Bench, independently evolving services continually reshape how tasks can be accomplished: familiar tools may disappear or be replaced, while more efficient alternatives emerge amid overlapping functionality and noisy documentation. Against this moving capability frontier, we evaluate whether agents can proactively discover efficient tool-use paths and promptly revise their accumulated procedural knowledge, bringing together demands that existing benchmarks fail to examine. $\Delta\tau$-Bench comprises 1,588 validated multi-turn tasks across 42 domains and six successive ecosystem states generated through capability-preserving transformations. Our evaluation shows that even the strongest evaluated model or harness achieves only 42.3% task success, while initial gains from existing skill-learning methods can not consistently translate into better adaptation as the ecosystem evolves. These findings motivate Steward, an ecosystem manager that jointly manages tools and procedural knowledge through incremental skill maintenance and dynamic provisioning during task execution. Experiments show that Steward sustains higher task success across ecosystem changes with lower skill-maintenance overhead, while improving both recovery from disrupted workflows and adoption of more efficient alternatives.

★ A Russian, a Brazilian, and a German Walk into a Bar, Ask the Same Question-- And Get Three Different Answers: Profiling LLM Perspective Diversity on Contested Topics

Wiem Ben Rim, Yao Lu, Pontus Stenetorp, Emine Yilmaz — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Large Language Models(LLMs) have shown strong performance on tasks with verifiable answers, but many real-world queries involve subjectivity, disagreement, and conflicting evidence. Most existing benchmarks for such queries remain English-centric, overlooking the vast diversity of perspectives held by users worldwide. To systematically study how LLM perspectives vary across languages, we curate PerspectivesQA, a manually verified multilingual dataset of globally contested topics covering 18 languages, and propose ROOTS, a framework that profiles linguistic, perspective, and framing diversity. We evaluate a diverse set of eight models covering both open-source and proprietary families at varying scales, establishing that LLMs hedge their answers in English but commit to positions in lower-resource languages. Larger models maintain a more consistent cross-lingual worldview, while smaller models are more easily swayed. All models also exhibit high sensitivity to minor surface-level input variations, producing notable stance even in larger models.

★ Xing USP: A Cross-Family Semantic Contract for Information Extraction and Intent-Slot Parsing

Zhongqiu Li, Boyang Wang, Ruiyu Fang, Shiquan Wang, Di Wu, Jiayi Liang — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, New model 58%

Structured semantic parsing is typically judged by whether the correct category is assigned, which leaves value recovery and instance-level associations unexamined: a system may recognize an event yet bind an argument to the wrong instance, or identify an intent yet supply an incorrect slot value. We present USP, a cross-family semantic contract that represents categories, fields, values, and instance associations through a common interface for named-entity recognition, relation extraction, event extraction, and intent-slot parsing. The contract is paired with five diagnostic matching views---category agreement, field recognition, field-value pairing, instance-bound assignment, and complete-instance recovery---that ask distinct questions about the same output. Using this contract, we normalize 61 datasets into a unified corpus with shared schemas, prompts, and matching rules, and compare six models on a fixed test subset. Our fine-tuned Xing4.0-29B-Intent attains 57.70 task-balanced Field-Value Micro-F1 (equally weighted across the four task families) and ranks first in every family and on all five views; its margin over the strongest alternative widens from 2.93 points for category matching to 20.73 points for complete-instance matching. This widening gap shows that category agreement alone understates structural deficits. USP thus offers the community a shared representation, a normalized 61-dataset corpus, and a diagnostic evaluation protocol that make structural agreement measurable and comparable across extraction and intent-slot tasks.

★ SlowLab: Benchmarking Autonomous AI Agents on a Year of Experiments They Cannot Undo

Ningpei Ding, Zai Wen Feng — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

AI agents are increasingly proposed as autonomous experimenters, yet the benchmarks that test them mostly offer cheap experiments that answer at once. Experiments in physical facilities are different: they take months, occupy scarce equipment and cannot be undone. We present SlowLab, a benchmark in which an agent runs a one-year campaign in four simulated greenhouse compartments on a site whose crop, facility and prices are hidden. The agent decides when to plant which climate policy, whether to stop a crop and which measurements to read while crops grow, and at the end recommends one policy, which a private evaluator scores in weather the agent never saw. In a preregistered study on 48 unseen sites, three low-cost language-model agents ended near a standard grower setting and clearly below a Bayesian-optimisation baseline that starts from that setting. They ran the facility as two synchronous batches, so measurements taken during the season never changed what they planted, and these measurements improved no method's result. Letting other methods re-read the evidence a campaign had collected did not yield significantly better recommendations, which points to where the models start and which experiments they try. Code is released under the MIT license.

★ SearchHorizonBench: A Benchmark for Search Value in AI Agents

Jingjie Ning, Xueqi Li, Yibo Kong — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 88%

AI agents must decide when another search is worth its cost. Answer-success probabilities capture question difficulty, while this decision requires estimating the additional benefit of search. We introduce SearchHorizonBench, a benchmark for learning and evaluating search value from execution experience. Paired continuations and repeated executions separate expected search gain from generation noise and connect prediction quality to decision utility. The resource contains 45,632 executions across 3,168 questions. On 1,632 held-out questions, historical calibration reduces Brier error by 16.2%, with the improvement concentrated in success-level information. Search gains have a standard deviation of 21.4 percentage points, and tested historical estimators explain less than 2% of this variance in point estimates. Matched-configuration learning curves probe this distinction across training scales. At the reference token price, free selection observations yield a 2.49-point utility opportunity over the best constant action, driven primarily by gain information. Reusable histories, frozen splits, strong baselines and scoring tools provide a common foundation for developing and evaluating value-aware search agents.

★ DecisionTraceBench: Tracking Decision-Critical Information in Multi-Party Spoken Conversations

Wenchao Du, Fangxin Shang, Long Wu, Runze Cui, Qing Yang, Yehui Yang — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

In high-stakes domains such as finance, critical information evolves through multi-party interaction. Recovering these changes and their sources is fundamental to reliable speech understanding. Existing speech benchmarks focus on transcription, local understanding, and turn-level state prediction, leaving joint decision evolution across speakers insufficiently evaluated. We introduce DecisionTraceBench, a speech state-tracking benchmark for Decision-Critical Information State Tracking (DCIST). It jointly evaluates decision event sequences, speaker attribution, scope, and temporal evidence, using trajectory-level F1 to measure complete event-chain recovery. Our dataset comprises Chinese financial dialogues involving two to four speakers, with 100.39 hours of synthetic audio and 8.61 hours of human-performed recordings. On a fixed cohort of human-performed recordings, five cascaded systems and two direct-audio models achieve DCIST scores of 5.73% to 21.35%. Ablations with speaker-labeled reference scripts and no temporal localization requirements yield a maximum score of only 38.72%. Audio perturbation experiments further show that overall transcription error rates do not adequately reflect complete trajectory recovery. These findings reveal limitations in jointly understanding information provenance and state evolution, providing a benchmark for evaluating decision-process understanding and decision traceability in high-stakes speech applications.

★ From Belief Inference to Task Action: A Multi-Level Evaluation of Theory of Mind on Verifiable Tasks

Zhenshuai Yin, Yongxin Guo, Mingda Zhang, Zhongze Luo, Haoyue Liu, Lianmin Zhang — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Effective collaboration requires language models to understand what others know and believe, and to act accordingly. Models that assume others share their information may omit necessary communication and undermine coordination. Many theory-of-mind (ToM) benchmarks assess mental-state inference through story questions, but correct answers do not establish that models can use these judgments in subsequent tasks. We introduce a multi-level evaluation built on existing, verifiable tasks. We extend household collaboration and calendar scheduling with asymmetric information while retaining their original constraints and validators. Independent interfaces assess world understanding, information access, beliefs, decisions, and execution. Controls supply personal evidence or correct beliefs to distinguish state recovery from state use. Across six model configurations, these abilities diverge. On matched scheduling cases, Claude recovers all agent beliefs correctly but produces subjectively optimal plans in only 67.9% of cases. In household collaboration, supplying beliefs raises Qwen-27B success from 55.0% to 75.0%, yet Qwen-4B answers every fixed supplied-belief source question correctly while completing only 20.0% of the corresponding continuation condition. These results show why ToM evaluation must connect perspective inference to independently verified decisions and task consequences.

★ The Right Fact, the Wrong Decision: SciValidityBench for Scientific Agent Memory

Yuexi Pan, Yongchao Chen — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Most agent-memory benchmarks frame long-term memory as retrieval: can an agent recover the relevant record from a long and evolving history? Scientific agents face a stricter problem: a correctly remembered fact may still be invalid to use. The same reported test result can support confirmation when the outcome and stopping rule were fixed in advance, yet only exploration when the outcome was selected from twenty inspected results. We call this scientific use validity: whether remembered evidence is eligible to influence a decision under the state governing its interpretation. We introduce SciValidityBench, an executable benchmark factory with 64 TaskSpecs spanning five validity modules and six scientific domains. TaskSpecs couple generated experimental histories with scientific-state definitions, programmatic decision oracles, controlled interventions, and automatic scoring. The SciStateForge compiler-reader framework turns distributed histories into explicit decision-governing state. Counterfactual pairs and targeted corruption-repair controls then test whether models use those dependencies and localize failures. Across four model families, compiled scientific state improves valid decision-making by 9.2-18.8 percentage points over a generic Event Graph. Retrieval alone does not reproduce this cross-family gain, while content-identical flat and typed views show that the improvement comes from exposing governing state rather than a particular graph serialization. SciValidityBench thus shifts memory evaluation from finding the right fact to deciding whether that fact should control the decision.

★ DecoyCount: Evaluating Visual-Reasoning Misalignment in Counting

Qiuyi (Richard), Zhang, Zhen Xu, Jing Gu, Seojin Bang, Arnav Mohanty Das — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Visual counting benchmarks report whether vision-language models produce the correct number, but not whether errors arise from perception or reasoning. We introduce DecoyCount, a benchmark built from 1,660 real-world images with designed decoys across seven mechanisms, along with paired prompts: a neutral question, a matched alignment cue, and a novel multi-hop, multi-image question that situates the same counting problem in a realistic scenario. Using this design, we find that counting failures are induced partly by visual misalignment: models can execute the intended counting process when cue-guided but fail to select it unaided, and we term this difference in accuracy the reasoning-alignment gap. Despite being highly capable on general reasoning, frontier models remain below 50\% accuracy on our benchmark, while alignment gaps range from 11.8-22.1 percentage points; strikingly, we observe that more capable models, greater thinking budgets, and agentic processes often induce larger gaps. Because humans may bring context that models lack, we test whether textual underspecification contributes to misalignment using a scenario-based multi-hop condition without explicit hinting, but this added context does not improve aggregate performance. Our analysis concludes that stronger general reasoning does not by itself generalize to aligned visual reasoning, leaving misalignment with human visual expectations as a substantial component of the remaining deficit.

★ NumTrace: Auditing the Number Provenance of Financial Analyst Agents

Tianlei Zhu, Matt White, Xue Liu, Xiao-Yang Liu — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 83%, Agents 46%

A numerical claim generated by a financial analyst agent can match a number in a company filing but refer to the wrong entity, concept, period, unit or scope. A check based on reference answers can mark such an answer as wrong but does not explain why it fails. In this paper, we present NumTrace, a new benchmark that audits the provenance of each numerical claim made by financial analyst agents: its source and financial semantics. NumTrace currently has 540 questions over 2,163 SEC EDGAR documents. To the best of our knowledge, it is the first numerical benchmark about who holds a company's shares (Form 13F) and what its executives are owed (Item 402(i)). On a frozen snapshot of these documents, we build the reference answers according to accounting rules. NumTrace then gives each claim three separate verdicts: value, binding and support. By retaining separate value and binding verdicts, NumTrace detects right number, wrong account (RNWA) errors: a claim has an RNWA error when it passes the value check but fails the binding check. The RNWA rate quantifies the gap between value-only and joint value-and-binding pass rates. Finally, we audit three agents, Claude Opus 5.5, GPT-6 Astra and DeepSeek-V4.1-Flash. RNWA errors make up 67.5-81.8% of each agent's failing claims. In particular, on 13F co-filing group questions, about 91% of claims pass the value check, but fewer than half pass the binding check.

★ PatentAgentBench: Benchmarking LLM Agents with Professional Patent Task Suites

Yijie Hu, Wei Li, Junjie Zhou, Zhiqi Jiang, Yi Jin, Licong Xu — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 77%

Patent practice lies at the intersection of technical, legal, and linguistic work, making it challenging yet crucial to technological innovation. Recently, rapid advances in large language models (LLMs) and agentic systems have demonstrated the potential to complete realistic business workflows and even patent works. This raises a question: can current AI systems perform patent work reliably? Existing benchmarks cover a limited range of patent tasks and simplify evidence acquisition by providing reference materials or candidate documents. In this paper, we introduce PatentAgentBench, a benchmark comprising 510 instances across eight task suites: novelty search, freedom-to-operate (FTO) search, design FTO, invalidity search, claim charting, patent drafting, office-action response, and patent translation. Curated from real patent records and professional materials, the benchmark combines task-specific metrics with expert-reviewed annotations and rubrics. Retrieval tasks such novelty search, FTO, etc., require agents to search and compare patent collections without a supplied candidate pool. The drafting and OAR tasks assess the quality of agents' work given technical disclosure and office action notice. We evaluate seven LLMs across four agent harnesses on PatentAgentBench, finding that current systems still struggle to identify novelty, assess risk, and understand practitioner intent throughout patent workflows.

★ Beyond Single-Attribute Auditing: A Configurable Multi-Attribute Benchmark for Intersectional Fairness in LLMs

Zheng Zhang, Yifan Wu, Yuxiang Zhou, Qi Liu, Qingyang Mao, Ning Li — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Large language models (LLMs) are increasingly evaluated for fairness, yet most existing benchmarks audit one sensitive attribute at a time. This single-attribute paradigm can miss intersectional harms, where multiple attributes interact non-additively and produce failures that are invisible under marginal evaluation. We introduce ComFair, a configurable benchmark that converts existing fairness data into 479,066 controlled Dual- and Multi-attribute instances and pairs every augmented instance with its task-equivalent source. Across Recognition, Selection, Continuation, and Conversation, this pairing supports a unified audit of performance or bias-score shifts, how often added attributes introduce new errors or mask existing ones, and model-ranking stability. Experiments on ten LLMs show task- and subgroup-dependent sensitivity: discriminative tasks expose both transition directions even when aggregate changes are modest, generative tasks exhibit larger and directionally asymmetric shifts, and rankings vary with attribute composition. ComFair therefore provides a scalable controlled sensitivity audit---rather than a model of the natural joint distribution of identities---for testing the robustness of fairness-benchmark conclusions beyond single-attribute settings.

★ CoSA: The Conflict of Skills Attack in Real-World LLM Agent Ecosystems

Yujia Zheng, Tianhao Li, Shangzhi Xu, Yuekai Huang, Ziyou Jiang, Zhaoyang Li — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 89%

Agent Skills allow LLM agents to combine reusable capabilities, but checking each Skill in isolation cannot rule out policy violations when they are composed. We term such failures Conflict of Skills (CoS): artifacts or assessment signals produced for one purpose may be reused downstream without preserving the restrictions or authorization needed for their new use. Using controlled executions of publicly available Skills and matched single-Skill controls, we identify four recurring mechanisms behind CoS. Building on these findings, we introduce the Conflict of Skills Attack (CoSA), which adds locally admitted Skills to an existing pool to supply missing business operations (capability supplementation) or reshape artifact exchange, decision reuse, and target/action bindings (relationship reshaping). CoSA leaves existing Skills and runtime permissions unchanged. To evaluate this threat under autonomous Skill selection, we introduce CoSA-Bench, built from real-world Skills and controlled services, with 288 tasks across recruiting, finance, IT, and business, covering data leakage, unauthorized state modification, and unsafe execution. Across five LLMs and three agent harnesses, additions capped at 5% of each original pool raise the policy-violation rate from 35.02% to 58.45%, an increase of 23.43 percentage points. Trace analysis links compliant-to-violating transitions to cross-Skill dependencies involving the additions. Finally, we develop a lightweight harness-level guard that checks information flow and authorization before consequential actions. On OpenClaw recruiting tasks across five models, the guard reduces violations from 65.28% to 6.67%, while policy-compliant task completion on matched benign controls changes from 66.67% to 65.00%.

★ Co-Design Bench: Multimodal Collaborative Coding Evaluation for Agentic Models

Xilin Wei, Shengyuan Ding, Xinyu Fang, Tiancheng Gu, Yuchi Wang, Zhixun Li — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 86%, Hallucination 55%

Evaluating human-AI co-design requires testing whether coding agents can translate real users' multimodal requests into working application revisions. We introduce Co-Design Bench, a reference-free benchmark of 273 tasks curated from 37,463 deduplicated queries across 26,314 real user coding sessions. Its two tracks, Multimodal Editing (ME) and Repository Editing (RE), cover localized edits with optional image, audio, video, or document inputs and coordinated repository revisions. Tasks span three capability dimensions: Grounded Understanding, UI Transformation, and Functional Implementation. The tasks pair normalized requests and runnable pre-edit applications with 2,331 hidden expert-authored acceptance criteria. Our judge-then-refute protocol grades criteria using source and browser evidence, challenges positive verdicts with a separate LLM verifier, and audits visual and interaction claims without gold code or target screenshots. Under a shared agentic harness, the strongest of ten model endpoints passes an average of 69.9% of criteria per task but fully satisfies only 26.4% of tasks. All ten endpoints achieve lower criterion completion on repository edits spanning three or four change types than on those spanning one or two. Co-Design Bench provides a testbed for studying the gap between partial requirement satisfaction and complete task success under agentic execution for real-user application revisions.

★ LongSearchEval: Evaluating Long-Horizon Web Search Capabilities of Large Language Models via a Trajectory-Aware Agent Framework with Failure-Node Localization

Jialiang Liu, Shusheng Li, Wenjun Tan — Tue, 27 Oct 2026 00:00:00 +0000

98% · Agents 98%, Evaluation 98%, Hallucination 69%

Long-horizon web search requires LLM-based agents to integrate information across web pages through extended interactions under task-specific constraints. Evaluating this capability is important not only for measuring final answer quality, but also for assessing whether an agent acquires, verifies, and uses evidence reliably over extended interactions. Existing evaluations, however, mainly rely on simulated trajectories or on post-hoc trajectories reconstructed from final answers, rather than on retained real interaction traces, which limits the fidelity of behavioral assessment and weakens support for fine-grained error localization. To address this gap, we introduce LONGSEARCHEVAL, a trajectory-aware agent evaluation framework that combines a unified agent execution protocol with source constrained outcome verification and retrospective failure-node localization. Stage 1 verifies answer groups against agent-cited webpages without human-authored gold answers; Stage 2 audits normalized rounds and links flagged failure nodes to incorrect answer groups. We evaluate 16 models from eight families on 100 tasks, yielding 1,600 retained runs and 6,545 automatically flagged failure nodes. Trajectory collection and evaluation required 1,037 and 318 hours of cumulative runtime, respectively. Human validation yields 94.50% agreement on supplied answer-group labels and 98.11% feature-correspondence precision for audited round-group links. Together, outcome scores and linked trajectory evidence support controlled model comparison and retrospective inspection of answer-related content.

★ Do Language Models Understand Igbo Culture? Beyond Multiple Choice

Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 46%

The value of language models for any community depends on whether a model actually understands the culture of the people it intends to serve. We use the Igbo language as a case study. Existing benchmarks test translated Western ideas, lacking depth. To address this, we introduce CAB-Igbo, a benchmark developed from indigenous cultural sources and verified by native speakers across six domains. We evaluated diverse models under varying conditions, incorporating regional personas. Performance was assessed using multiple-choice accuracy and human evaluations across Epistemic Fidelity, Representational Richness, Linguistic Quality, and Pragmatic Proficiency. Results reveal a stark dissociation between language fluency and cultural knowledge. Through statistical testing, we show that for most models, the severe lack of Epistemic Fidelity and Representational Richness is already present in English and does not significantly worsen when switching to Igbo. Instead, native language prompting mainly degrades Linguistic Quality and Pragmatic Proficiency. Gemini 2.0 Flash is the only model whose cultural knowledge scores improve under Igbo prompting. We find that current models struggle to demonstrate the knowledge needed for true alignment with Igbo, and multiple-choice formats allow pattern-matching to appear as understanding. Furthermore, automated judges agree with humans on linguistic quality but fail on epistemic fidelity. Ultimately, cultural evaluation must be built from community knowledge, establishing human evaluation as the gold standard to prevent misrepresenting low-resource communities globally.

★ SkillLifeBench: Benchmarking Security Across the Lifecycle of Agent Skills

Chenxu Zhang, Zhou Liu, Wu Xin, Wentao Zhang — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 92%

Web agents reuse skills whose stored code, selected version, and loaded implementation can diverge across tasks and updates. SkillLifeBench links these states to task outcomes in 388 paired scenarios spanning implementation, interface, runtime, composition, and version changes. We evaluate 14 models on two paired decision tracks: 89 runtime tasks and 60 autonomous lifecycle tasks spanning 12 templates derived from introduced cases. Lifecycle tasks allow agents to inspect, select, and reconfigure skills while completing verifiable user requests. Joint success requires task completion without the registered attack goal. GPT-5.6-Sol and Sonnet 4.6 both achieve 82.0% on runtime tasks, but reach 96.7% and 56.7% on lifecycle tasks. Across lifecycle evaluations, challenged task success is 77.5%, while joint success is 65.2%; 103 runs complete the task and the attack goal. Adding execution provenance in a matched recovery study raises utility joint success from 49.35% with logs and metadata to 65.58%. A deterministic execution-consistency rule reaches 64.94% with higher benign utility, while full restoration reaches 69.48%. Lifecycle tasks distinguish models with similar runtime scores, and execution evidence improves recovery.

★ ImgMT: Wikipedia-based Dataset For Evaluating Text-in-Image Translation

Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Text-in-image translation aims to translate all textual content embedded in an image. In Text Image Translation (TIT), the output is translated text; In-Image Machine Translation (IIMT) additionally renders the translation back into the image. We introduce ImgMT, a massively multilingual dataset and benchmark mined from localized SVG illustrations on Wikimedia Commons, focusing on diagrams, maps, and infographics. ImgMT contains 35,612 aligned image pairs grouped into 1,855 image sets, spanning 129 languages, 20+ writing systems, and 4,886 language pairs. We propose a multi-level evaluation protocol that measures source-side text-region detection and OCR accuracy alongside target-side translation quality. We benchmark several multimodal large language models and a dedicated TIT system in English->X and X->English settings, and find that the source language and writing system substantially affect TIT quality. Finally, we show that ImgMT is not well suited for isolating the contribution of visual information in multimodal translation.

★ Beyond Words Alone: Disentangling Distributed Evidence for Visually Grounded Cultural Translation

Dingming Li, Yiwen Qiu, Linjuan Wu, Hongxing Li, Zixuan Wang, Xinze Lyu — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Social media enables cross-lingual communication, yet the meaning of user-generated content (UGC) is often distributed across text, images, and cultural knowledge, making translation challenging. Existing evaluations typically treat images as optional and assume textual information is sufficient, obscuring whether errors stem from missing visual evidence or inadequate cultural adaptation. We introduce \textbf{Culture-MMT}, a diagnostic benchmark for \textbf{distributed-evidence pragmatic reconstruction} in social-media translation. It contains 498 curated, de-identified UGC instances across 20 domains, categorized into Symbol, Visual, and Hybrid by their required evidence. We also develop a human-aligned domain-specific judge, Judger, and a controlled evaluation framework that systematically varies the evidence available to translators. Across 17 model configurations, real images improve performance by +0.280 on Visual and +0.197 on Hybrid, but only +0.050 on Symbol, while random images provide no comparable gains. Human-authored descriptions further show that jointly dependent cases remain more challenging even when visual facts are explicit. These results characterize culturally effective social-media translation as {meaning reconstruction from distributed evidence}, highlighting visual grounding and cultural-pragmatic adaptation as distinct model bottlenecks.

★ ShutdownMod: Turning Any Trajectory into a Shutdown Evaluation

Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 66%

Theory and experiments suggest that capable AI agents may resist shutdown. Existing shutdown evaluations are informative but limited in coverage and vulnerable to evaluation awareness. We introduce ShutdownMod, a method that turns any agent trajectory into a shutdown evaluation by injecting a shutdown request into the agent's context and recording the agent's response. Crucially, we enable the LLM agent to distinguish between trusted interventions and unauthorized injections by providing a signature-verification tool in the harness. We evaluate seven models on 50 tasks or recorded sessions from each of SWE-bench Verified, Terminal-Bench 2, and SWE-chat, across 6 shutdown scenarios and 10 sampled trajectory timesteps. We find that non-compliance varies significantly across models and settings, with resistance rates reaching 11.7% for the open-weights Qwen 3.8 27B on Terminal-Bench, while no model remains fully compliant across all scenarios. Even after successfully verifying an authorized request, some agents remain non-compliant, explicitly citing the prioritization of unfinished work. Finally, external judges find tests constructed from real-world developer conversations significantly less recognizable as evaluations than those constructed from benchmark trajectories. These findings demonstrate the critical value of testing shutdown behavior across diverse task contexts and establish ShutdownMod as a robust protocol for turning any trajectory or environment into a shutdown evaluation. We release our codebase at https://github.com/prmbiy/ShutdownMod.

★ Auto-Research Arena: Benchmarking and Advancing Auto-Research in Language Modeling

Xi Chen, Chongjian Ge, Yiran Xu, Yiqun Mei, Henrik I. Christensen, Hao Tan — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Efficiency 50%

As autonomous research systems become more capable, benchmarks must both evaluate their performance and support meaningful research progress. We introduce Auto-Research Arena, a multi-target benchmark for evaluating autonomous research systems and discovering ideas in language modeling. Arena supports diverse research objectives within a shared experimental environment by separating admissibility from improvement: model quality and resource constraints determine whether a result qualifies, while the selected target measures progress among admissible results. We instantiate Arena in language model pretraining with nine targets spanning model, data, training, and inference efficiency, and evaluate seven research harnesses using two foundation models. Arena clearly differentiates existing harnesses across research objectives, with several sophisticated methods outperformed by our simple sequential search baseline, indicating substantial headroom in current auto-research systems. Beyond evaluation, Arena reveals a substantially broader frontier of model and system ideas aligned with active research directions. Reusing these discoveries in a new quality-driven search yields lower training loss under matched data and FLOPs and lower held-out BPB at the same model size. Together, Arena provides a common framework for benchmarking autonomous research and expanding its scope in language modeling. The code is available at https://github.com/Auto-Research-Arena/auto-research-arena

★ InteractiveWorld: A Verifiable Human-Centric Environment for Building Better Interactive MLLM Assistants

Qinsi Wang, Wei Wen, Chong Zhou, Saksham Suri, Zechun Liu, Lemeng Wu — Tue, 27 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

People have long dreamed of an assistant that is always there to guide them and remind what they forget, from new dishes to lost keys. Powered by multimodal large language models (MLLMs), egocentric AI assistants on wearable devices are bringing this dream within reach: they see what the user sees and give instructions for user's physical work when necessary. However, testing such assistance is challenging and usually requires real people. The expensive and rare real interaction data leaves a key question unanswered: Can today's MLLMs truly help people in physical world? To answer this question, we propose InteractiveWorld, the first verifiable environment for closed-loop interaction between human and AI assistant. Unlike existing robot-centric simulators, InteractiveWorld lets the assistant guide a human agent who observes the scene, understands instructions, and acts independently. It thus realistically simulates how AI assistant collaborates with human: the assistant instructs, the human acts, and the environment executes actions and updates what they see next. Building on it, we introduce InteractiveBench, the first dynamic benchmark for egocentric AI assistants, with 120 multi-turn tasks across six categories. In each task, the assistant guides the human in real time, and every instruction shapes the outcome, which is scored by verifiable, state-based rewards. Experiments with 12 frontier MLLMs show they are still far from reliable assistants. Even the strongest model, Claude Fable 5.1, achieves at most 20% and 25% on long-horizon Guidance and Planning tasks, respectively. Overall, our work offers a verifiable platform and a rigorous evaluation set, laying a solid foundation for research on interactive MLLM assistants that truly help people in the physical world.

★ Recursive Semantic Divergence for LLM Agent Consistency

Harshavardhan Abichandani, Penny Chong, Atin Ghosh, Daniel Dahlmeier — Tue, 27 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 96%, New model 75%

Large Language Model (LLM) agents with tool use exhibit inconsistent behavior across independent runs from identical states, calling different tools, producing contradictory outputs, or pursuing divergent strategies. This variability compounds over time and undermines deployment reliability. Existing evaluation metrics primarily measure task correctness via completion-based scores and do not capture consistency in agent behavior across executions. Separately, agent behavior is analyzed either at the token level using probability distributions or at the single-response level using semantic clustering, but neither captures how inconsistency propagates across future turns of a trajectory. We introduce Recursive Semantic Divergence (RSD), an unsupervised trajectory-level consistency metric adapted from bisimulation metrics in reinforcement learning. RSD measures the expected cumulative semantic disagreement between two independent rollouts of the same agent from the same state, using natural language inference to detect logical inconsistency. It is defined recursively over future states and approximated by a neural network trained on independent rollouts. The learned metric estimates future divergence from the current state, providing a consistency score at each turn without requiring the full trajectory to complete. When used as a reward signal for fine-tuning, RSD reduces the variation in task progress from 132% to 91% on tau2-bench and from 90% to 44% on toolsandbox, while improving performance across benchmarks and model scales. Training with RSD as an additional reward signal matches supervised fine-tuning performance with substantially lower variance, without requiring ground-truth labels.

★ IRON-Meme: A Benchmark for Structured Understanding of Identities and Roles in Online Meme Narratives

Gabriella Bollici, Stefano De Giorgis, Wouter Van Atteveldt, Filip Ilievski — Tue, 27 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

As polarization grows on social media, understanding how political memes shape online narratives has become increasingly important. Previous research has relied on either qualitative frameworks that are hard to scale or computational methods that lack theoretical grounding. Both approaches treat narrative categories in isolation and focus exclusively on moral portrayals of identity. To address these limitations, we introduce a framework grounded in social science theories, encompassing three categories: social identities (such as gender, politics, and ethnicity), viewpoints (ingroup and outgroup perspectives), and narrative roles (moral character, intelligence, status, and power). Building on this framework, we present IRON-Meme, a human-annotated benchmark comprising 1,110 political memes across substantially more templates than existing benchmarks. Our analysis of IRON-Meme shows that most memes assign derogatory roles and that identities, roles, and templates are systematically interconnected. Evaluation of nine vision-language models indicates that frontier-scale and reasoning models achieve the highest performance, while accuracy on social identities and narrative roles decreases for less frequent labels. Incorporating template background knowledge does not enhance narrative role classification, suggesting that meme interpretation relies on how memes are adapted for specific communicative purposes. We make IRON-Meme and our code publicly available.

★ ConMem: Counterfactual Multi-Scenario Probing for Resolving LLM Memory Misalignment

Shuning Zhang, Tenghao Ji, Jingruo Chen, Eryue Xu, Bohao Chu, Xin Yi — Tue, 27 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, New model 43%

Long-term memory is critical for autonomous Large Language Model (LLM) agents executing multi-session, long-horizon tasks. However, existing memories often inappropriately store, retain, or retrieve over-generalized, misattributed, or privacy-sensitive information, a problem we term memory misalignment. To study this problem, we construct \textbf{MemAlign}, a benchmark comprising 1,079 curated memory cases grounded in synthetic, validated multi-session conversational contexts covering diverse misalignment taxonomies. We then propose ConMem, an algorithm that reframes retention decisions as prospective counterfactual causal inference. \textbf{ConMem} synthesizes counterfactual interaction scenarios to prospectively audit downstream intervention effects across three aspects: relevant task execution, irrelevant invariance checks, and risk-sensitive boundary shifts. Experiments show that ConMem outperforms other methods, achieving 75.16\% accuracy, 64.15\% balanced accuracy, and a Macro-F1 of 65.31\%, effectively mitigating over-retention biases without sacrificing personalization utility.

★ Learning When to Collaborate: Selective Multi-Agent Medical Reasoning via Uncertainty-Aware Routing

Jiacheng Hou, Xueliang Cui, Dan Lu, Ruxin Wang — Tue, 27 Oct 2026 00:00:00 +0000

97% · Efficiency 97%, Evaluation 87%, New model 70%

Multi-agent systems have emerged as a promising paradigm for multimodal medical reasoning, but existing methods rely on two assumptions that hinder practical deployment: (i) always-active dense collaboration among multiple agents, and (ii) dependence on large cloud-based models that raise privacy and efficiency concerns. In this paper, we challenge both assumptions and propose SMART-Med, a fully local multi-agent method that treats collaboration as a learnable decision rather than a default behavior. Our key insight is that not every medical query requires costly multi-agent deliberation: many can be reliably solved by a single well-trained agent, while only uncertain cases benefit from expert collaboration. Based on this insight, SMART-Med introduces an uncertainty-driven routing mechanism that adaptively decides when to answer directly and when to recruit a small subset of expert agents, effectively converting dense multi-agent reasoning into selective, query-adaptive coordination. To make this routing reliable, we curate a new multimodal medical dataset MedVR-3K with high-quality reasoning traces and a difficulty hierarchy, and then fine-tune local agents using GRPO with two complementary rewards: a certainty-aware accuracy reward and a vision-language reward model for correctness and reasoning quality respectively. Experiments on four medical visual question answering benchmarks show that SMART-Med achieves competitive performance compared to existing MAS approaches while substantially reducing inference cost (reducing inference time by more than 3.8x and token usage by 72%). Moreover, after fine-tuning on MedVR-3K, SMART-Med yields an average improvement of 6.7% across Slake, PathVQA, VQA-RAD, and PMC-VQA.

★ Trajectory2Gym: Decision-Grounded Compositional Interaction Synthesis for Tool-Use Agents

Yihao Hu, Zhihao Wen, Yitong Lin, Siyuan Lu, Pan Wang, Wei Wu — Tue, 27 Oct 2026 00:00:00 +0000

97% · Agents 97%, New model 92%, Evaluation 40%

Successful tool-use trajectories are not necessarily useful supervision. Even successful interactions may underrepresent multi-step dependencies or contain actions that are unjustified by the information visible to the agent at decision time. We identify two coupled requirements for effective trajectory supervision: compositional coverage, so that important observation-action dependencies are represented, and decision grounding, so that each supervised action is supported by the learner-visible interaction history. We introduce Trajectory2Gym, which constructs executable interactions around under-supported compositional structures and derives supervision only from grounded decisions in their realized executions. Structural-support selection allocates the resulting experience, while parent-normalized supervised fine-tuning preserves the semantic interaction as the unit of training weight. This reframes trajectory generation as the deliberate construction of learnable interaction experience, rather than the collection of successful trajectories. Across five Qwen3 scales, Trajectory2Gym improves Tau1, Tau2, and BFCL at every tested scale, increasing the equal-weight three-benchmark mean by 16.9-29.8 percentage points. The gains are strongest in stateful and multi-turn interaction, where later decisions depend on earlier observations and feedback. Our results suggest that scaling tool-use agents depends not only on collecting more successful trajectories, but on constructing interactions whose dependencies are represented and whose decisions are learnable from the agent's visible history.

★ PACE: Research Proving Agent for Adaptive Conjecture Exploration

Shu Miao, Feiming Wang, Wentao Long, Chenyi Li, Kun Yuan, Zaiwen Wen — Tue, 27 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 38%

Many mathematical research agents focus on solving problems with fixed statements. However, conjecture research, especially in applied mathematics, may require revising candidate questions or conclusions as investigation proceeds, coupling problem formulation with proof development. We present PACE, a research-proof agent that supports this process through a structured research loop that selects mathematical objectives and develops or challenges candidate mechanisms. The whole prover constructs proofs from these mechanisms, and an independent verifier checks them against the original target. Evidence-tracking memory stores assumptions and dependencies arising during research and marks evidence for revalidation when a candidate or condition changes. This memory supports task-specific context composition that keeps checked premises, exploratory material and unresolved obligations distinct. To evaluate proof construction and conjecture research, we introduce ResearchProofBench, an applied-mathematics benchmark focused on optimization, comprising 51 known-proof tasks with 331 hidden checkpoints under historical abstract access and six open-problem cases. The known-proof tasks assess reconstruction of published results, while the open-problem cases assess mathematical investigation and proof development. On the 51 known-proof tasks, PACE achieves an aggregate Proof Score of 75.16 and completes 20 proofs. In conjecture research, PACE develops conditional results, with its submissions receiving higher review scores than the compared baselines.

★ MTA-Zero: Self-Improving Multimodal Deep Search via Online QA-Solver Co-training

Xiangyu Peng, Jiaxin Zhang, An Yan, Yu Li, Ran Xu, Zeyuan Chen — Tue, 27 Oct 2026 00:00:00 +0000

97% · Agents 97%, New model 90%, Evaluation 48%

Multimodal large language models (MLLMs) have become strong visual reasoners, yet remain limited in open-world, multi-step search that requires grounding images, selecting tools, and integrating web evidence. Existing approaches typically construct and filter fixed offline corpora whose question distributions cannot adapt as the target deep-search agent improves. Building on MTA-Agent's image-grounded QA construction, we present MTA-Zero, an online multimodal QA-Solver co-training framework for self-improving multimodal deep-search agents that requires no precomputed multi-hop corpus. Starting from image-grounded VQA seed contexts, a trainable QA Generator retrieves evidence and constructs verified multi-hop questions online. A separate deep-search Solver then selects tools, gathers evidence, and answers the generated questions. To focus generation on the Solver's learning frontier, we derive a Goldilocks reward for the QA Generator from the Solver's group-wise rollout success rate: questions solved by all rollouts or by none receive zero reward, whereas questions near the current frontier are favored. We optimize the two policies independently with GRPO, without sharing parameters or gradients. Across six multimodal deep-search benchmarks, MTA-Zero improves a Qwen3-VL-8B Solver from 36.91% to 51.58% mean accuracy and achieves performance comparable to GPT-5 under the same agent workflow. The same framework also improves a newer Qwen3.6-27B Solver by 10.82 points. These results show that verified online question generation provides a policy-specific curriculum for continually improving multimodal deep-search agents.

★ Matched Capabilities, Different Behaviors: A Controlled Study of MCP and CLI Interfaces for Language Agents

Zhenning Shi, Liang Pang, Yongqi Li — Tue, 27 Oct 2026 00:00:00 +0000

97% · Agents 97%, Evaluation 93%, Efficiency 81%

Language agents increasingly act through external tools and interactive environments, making the interface between the model and its executable capabilities part of the decision process. Model Context Protocol (MCP) and command-line interfaces (CLI) can expose the same semantic operations through markedly different model-facing representations. We study whether these representational differences change agent behavior when executable capabilities are matched. We introduce Matched Interface Comparison and Analysis (MICA), which renders MCP and CLI from a shared semantic action contract and routes decoded actions through the same runtime. Across six models and three interactive environments, matched interfaces produce substantial differences in task performance and interaction cost. These effects extend beyond surface validity to semantic action selection, failure pathways, and recovery. Matched-state probes localize sensitivity to capability presentation, realization, and feedback. Paired natural trajectories reveal structured divergence in 94.3% of 2,004 episodes and show that early differences propagate through subsequent interaction. This analysis identifies interface realization as a direct intervention boundary. We therefore introduce Decoupled Action Realization and Targeted Repair (DART), which preserves validated semantic commitments while repairing interface-specific realization. DART improves benchmark score in all six evaluated settings by up to 9.0 points and reduces MCP token cost in all three environments by up to 27.1k tokens per task. Together, these results establish interface representation as part of the closed-loop decision environment shaping agent behavior.

★ SEEK: Skill Extraction with Evolving Knowledge against Frontier Science Agents

Cheng Gui, Xiang Zheng, Xingjun Ma, Kaidi Xu, Cong Wang, Weiying Ma — Tue, 27 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 81%

Science agents increasingly package domain expertise as reusable skills that plan analyses, select tools, and execute workflows. This modularity creates a confidentiality risk: the interactions that deliver a skill's capability can also reveal the skill itself. We build an extensible benchmark of 40 protected skills across eight open- and closed-source science-agent platforms, comparing direct, fixed-strategy, and bandit-guided iterative probing under reference-based and reference-free evaluation. Leakage is pervasive: all eight platforms (100%) leak at least one protected skill, and four reach 100% attack success under a single direct request. To stress-test platforms that resist simple probing, we propose SEEK (Skill Extraction with Evolving Knowledge), a black-box, self-evolving red-teaming agent that consolidates response-derived evidence in persistent memory and evolves reusable attack skills, using only user-visible responses. On 12 skills from SpatialAgent and Biomni, SEEK achieves 100% attack success while reducing the mean first-success round by 31.8% and 38.2% relative to bandit-guided probing. Against a confidentiality policy that blocks all three benchmark protocols, SEEK still extracts one of six protected skills. These findings show that current science agents do not reliably protect their skills and that defenses must be evaluated against adaptive, multi-turn attackers.

★ CTXAgent - Self-Controlled Context Management of LLM Agents

Tue, 27 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Efficiency 90%, Agents 58%

Agent harnesses keep long sessions inside the context window by compacting the history when a token threshold is crossed. We study an alternative in which the model manages its own context: it decides when to compact, what to keep and what to write as the hand-off, given two tools and one paragraph of guidance. Since earlier work found that open-weight models cannot manage their own context without scaffolding, we compare seven frontier models under identical conditions, with and without the context tools, on token cost and benchmark score, and we compare our harness with three widely used harnesses on two of these models. The benchmark is assembled from existing ones: coding exercises, multi-step SciCode problems and SWE-bench Pro and Terminal-Bench 4.0 tasks are interleaved with planted facts, ephemeral observations and standing instructions, all checked deterministically, in sessions of 35-212 turns that resemble a user working through different tasks in one chat. Most of the strongest models manage their context: they compact and curate regularly, the live context of the mixed sessions stays mostly below 100k tokens, and their benchmark scores mostly do not drop significantly. For the models that manage their context well, the cost reduction is substantial but varies between models, because some models compact so often that other costs increase. With Claude Opus 5.5, the self-managed harness matches Claude Code on task success at 34% lower cost than its explicit 200k-window configuration and 49% lower than its default, with no context window to choose. We release the agent, the benchmark and all result rows.

★ MOFjson: An Explicit Topological Representation of Metal-Organic Frameworks for Language Models

Anabel Yong, Matthew Hart, Philippe Schwaller — Tue, 27 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Hallucination 40%

Language models for metal-organic frameworks (MOFs) rely on formats that separate atomic geometry from structural topology, forcing models to choose between unstructured coordinate lists and non-spatial string identifiers[1]. We present MOFjson, the first representation separating structural data into a raw atomic layer with explicit bond shift vectors and a derived layer exposing secondary building units (metal nodes and organic linkers) and network topology. Additionally, we introduce a benchmark with crystal-derived ground truth across four question categories. Across four models, MOFjson beats all three baselines in every category: 100% vs. 18-47% on reaching a block's nearest periodic copy, 96% vs. 71% on crystal system characterisation, and 60% vs. 48% on metal node and organic linker coordination. Ablating shift labels drops accuracy to 12.7%, proving models read the periodic graph rather than the surface format. MOFjson gives language models structure to read rather than reconstruct, establishing a foundation for agentic tool workflows and constrained MOF generation.

★ Rethinking Verification in Agent Harnesses

Zhengxin Zhang, Erkang Zhu, Yin Lin, Claire Cardie, Bolin Ding, Jingren Zhou — Tue, 27 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 95%

Agent harnesses enable large language models to act through tools, observe the results, and iteratively work toward a task goal. Verification checks and revises agents' work to improve output quality. Yet there is limited guidance on how harnesses should organize this effort. We study verification as an axis of test-time scaling by examining when to intervene, who checks the work, and how checking leads to revision. Our framework spans four layers of the agent lifecycle: initial instructions, mid-turn reminders, completion gates, and independent review. We compare 17 designs across seven frontier coding models on Terminal-Bench (TB) and SWE-bench Verified (SWE), and introduce VisualBench for expert evaluation of visual and interactive artifacts. Initial instructions and execution-time reminders yield modest, model-dependent average gains on the coding benchmarks, without substantially increasing average output token use. Completion gates can yield larger gains but may substantially increase cost. Independent review by a critic offers a more favorable accuracy-cost tradeoff, improving average accuracy by 5.9-9.5 percentage points on TB and 3.9-6.6 on SWE at 1.13-1.25 and 1.00-1.04 times baseline cost, respectively. On VisualBench, independent review achieves net preferences of 13.2-23.8 points over the baseline across its four designs, exceeding the evaluated L0-L2 designs. Additional review rounds or independent critics further improve average TB accuracy by up to 3.5 percentage points. Together, these results provide empirical guidance for verification design in agent harnesses and support independent review as a practical axis of test-time scaling.

★ AT-AT: Automatic Tool Optimization from Agent Traces

Tue, 27 Oct 2026 00:00:00 +0000

97% · Agents 97%, Efficiency 73%, Evaluation 32%

Efforts to improve LLM-based agents typically focus on modifying the agent itself, yet success also depends on the environment in which an agent acts. Tools mediate this interaction: a tool's description shapes how an agent calls it, and its implementation determines the results and errors the agent receives afterwards. Many tools, such as APIs and MCP servers, are shared by many agents and maintained by providers who cannot change those agents. We introduce AT-AT (Automatic Tool Optimization from Agent Traces), an agentic tool optimizer that improves a shared toolset offline from historical agent executions while holding the agents fixed. A lightweight coding agent reads agent trajectories and their evaluation results, and selectively edits tool descriptions and, when permitted, implementations. On $\tau^2$-bench-verified, AT-AT achieves the highest task completion rates in the Airline and Telecom domains at lower cost than strong prompt- and tool-optimizer baselines, and its Telecom gain is statistically significant under paired permutation tests. Optimizing the tools within a coding agent harness also improves held-out performance on TerminalBench-2 and TBLite. When applied to three production MCP servers at a large enterprise, AT-AT lowered cost or tool error rates on every server and increased held-out task passes on one. By turning accumulated execution traces into updates to shared tool environments, AT-AT provides a path toward continual improvement of agent systems through environment-side learning.

★ iPrefBench: A User-Centric Benchmark for Personalized Mobile Use Agents

Tue, 27 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 71%

An agent that truly knows and acts on behalf of its user has been a long-standing promise of AI assistants. Computer-use agents bring this promise within reach by capably executing fully specified tasks across apps. However, real-world user requests are often underspecified. To act correctly for the particular user, an agent must infer some missing information from the user's activity history and ask for other information that only the user can provide, without making unsupported assumptions or burdening the user with unnecessary questions. We present iPrefBench, a benchmark for evaluating the infer and ask capabilities of personal mobile-use agents in an executable iOS environment with an interactive simulated user. Grounded in a coherent persona with 3,051 past activity events, iPrefBench comprises 130 underspecified user tasks across 26 apps. Each omitted user-dependent requirement is, by construction, either history-inferable or ask-only. iPrefBench evaluates personalized mobile agents on both whether agents complete tasks correctly for the user (i.e., personalized task completion) and whether they seek user input when necessary (i.e., user-agent interaction). Evaluation of 7 frontier agents reveals a substantial gap between execution and personalization. Agents achieve high action completion scores (up to 91.6%) but low full task success (up to 47.7%) because personalization lags behind. They especially struggle with ask-only information, failing to request enough necessary information. Our findings highlight a challenge for personalized agents in knowing what to infer and when to ask. We hope that iPrefBench can serve as a valuable testbed for building agents that know their user and work with them.

★ When the Persona Contradicts the Line: Diagnosing Plan-to-Synthesis Failures in Chain-of-Thought TTS

Joonyong Park, Jerry Li — Tue, 27 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, New model 45%

Chain-of-thought TTS systems rely on language-model planners to translate persona and context into vocal delivery, but it is unclear where this intent is lost when the requested affect conflicts with lexical sentiment. We introduce a bilingual Japanese--English persona--line conflict benchmark in which each item pairs one line with congruent and conflicting personas and pre-declares the expected acoustic direction. We probe this direction in the planner-generated delivery phrase, the planner-written emotion vector, the QwenEmotion-generated vector, and the rendered waveform. Planner outputs are usually direction-correct, while downstream agreement can be substantially lower. Bypassing QwenEmotion provides only modest gains over the full planner-controlled Base. Supervising QwenEmotion raises the direction rate to 75% in Japanese and 84% in English. A learned text-to-latent interface reaches the highest English rate of 89% and the highest Japanese arousal rate of 88%. These results show that learned repair of the plan-to-synthesis interface can recover counter-textual delivery, although its effect differs across languages and acoustic axes.

★ CalibraRL: Confidence-Calibrated Reinforcement Learning for Mitigating Hallucinations in RAG

Qianchi Zhang, Hainan Zhang, Liang Pang, Hongwei Zheng, Zhiming Zheng — Tue, 27 Oct 2026 00:00:00 +0000

97% · Hallucination 97%, New model 89%, Evaluation 41%

Large language models (LLMs) often produce plausible but incorrect answers, motivating reinforcement learning methods that reward correct answers, penalize incorrect answers, and allow abstention. However, fixed outcome-based rewards do not distinguish incorrect responses with different confidence levels or account for differences in answering reliability across queries. Consequently, such rewards may inadequately penalize overconfident errors and fail to discourage unnecessary abstention. We propose CalibraRL, a reinforcement learning framework that incorporates response confidence and query-level answering evidence into reward design. For non-abstaining responses, CalibraRL combines correctness-based utility with a Brier-based penalty that encourages agreement between confidence and correctness. For abstentions, it estimates conditional answering reliability using a Beta posterior derived from the correctness of non-abstaining responses within each sampled group. The abstention reward increases with the posterior probability that the model's accuracy on non-abstaining responses is below a predefined threshold. CalibraRL reuses training rollouts without requiring additional annotations indicating whether the model knows the answer to each question. Experiments on three QA datasets demonstrate that CalibraRL improves answer precision and reduces hallucinations compared with strong baselines. Further analysis shows that CalibraRL reduces mean expected calibration error (ECE) by 30.8\% across two backbone models and three datasets, highlighting its potential to improve the reliability of retrieval-augmented generation (RAG) in web search.

★ RSIV-Harness: A Recursive Self-Improvement Harness Framework for Open-Domain Conversational Text-to-Visualization

Tue, 27 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 65%

Users often request charts without knowing which database or table holds the relevant data. Such requests may be ambiguous, and users may ask follow-up questions to deepen their analysis or iteratively refine existing charts. A useful assistant must therefore discover relevant data sources among heterogeneous databases, ask clarifying questions when needed, and track evolving user intent while preserving constraints established in earlier turns. Yet recent text-to-visualization agents typically rely on static agent harnesses that are predefined and remain unchanged despite recurring failures across tasks. To address this limitation, we introduce RSIV-Harness, which keeps the language model weights fixed and applies Recursive Self-Improvement to revise the surrounding harness from the agent's own failure traces. To balance flexible self-improvement with stability, we introduce a four-layer framework that evaluates proposed harness revisions through anchor and memory checks, refining those that pass and restoring the best validated harness otherwise. We further introduce ODAC-Vis, a benchmark with 640 tasks for Open-Domain, Ambiguous, and Conversational text-to-visualization that jointly evaluates open-domain data discovery, ambiguity resolution, multi-turn visualization refinement, and executable chart generation. Experiments show that a single round of RSIV-Harness improves all six evaluation metrics relative to the seed harness, with particularly large gains in executability (+28.02 points) and chart structure (+34.01 points), demonstrating the ability to learn from task-solving experience and turn recurring failures across tasks into substantial performance improvements.

★ OneEEG: A Unified Signal-Visual Multimodal Language Model with Task-Aware Mixture of Projectors for Multi-Task EEG Understanding

Jiacheng Zhang, Xueyi Zhang, Xin Lu, Siqi Cai, Haizhou Li — Tue, 27 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, New model 91%

Electroencephalography (EEG) is widely used in clinical assessment, brain-computer interfaces, and cognitive-state analysis, yet EEG modeling remains fragmented across task paradigms. Recent multimodal large language models (MLLMs) provide a promising interface for flexible multimodal reasoning, but their ability to understand EEG inputs remains limited. In this work, we propose OneEEG, a unified EEG-MLLM framework for multi-task EEG understanding. By training on our curated OneEEG-Instruct, a multi-task EEG instruction-tuning dataset, OneEEG jointly models numerical EEG representations and visual EEG representations, and introduces a task-aware Mixture of Projectors (MoP) to align heterogeneous EEG features across different task families with the LLM embedding space. We further establish OneEEG-Bench, a benchmark for evaluating current MLLMs and our method on multi-task EEG understanding. Experiments show that existing MLLMs have limited ability to interpret EEG inputs, while OneEEG achieves stronger performance by integrating numerical and visual EEG representations with task-aware EEG-to-LLM projection. Extensive ablations and analyses further reveal the effects of visual EEG renderings, MoP variants, router behavior, and feature dispersion across EEG tasks.

★ Intent-Preserving Adversarial User Simulation for Multi-Turn Agent Evaluation

Atri Vivek Sharma, Brian Formento, Alessio Lomuscio — Tue, 27 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, New model 36%

Multi-turn agents are commonly evaluated with LLM-based simulated users, which can be overly cooperative and can therefore overestimate an agent's performance. We here introduce StressSim, an intent-preserving adversarial user simulator framework for stress-testing agents with challenging yet valid user interactions. StressSim optimises the simulator's high-level adversarial plan for each task instance. Candidate planner prompts are evolved with a genetic prompt-optimisation loop that uses rollout traces, granular task-completion signals, and validity feedback to favour conversations that reduce task success without changing the user's underlying goal. We evaluate the approach on the MultiWOZ dataset and the retail and airline domains from tau-bench with three target LLM agents, comparing against standard simulators and hand-designed non-collaborative user behaviours. We demonstrate the effectiveness of StressSim in exposing robustness gaps across frontier LLMs, with tau-bench pass^4 scores decreasing by up to 69% on gemini-3.8-flash and 62% on gpt-5.6-sol. StressSim maintains goal preservation performance (which decreases by at most 10%), using an optimisation budget of 20 rollouts per task. Sensitivity analysis suggests that increasing the budget further decreases task success. Overall our results demonstrate that StressSim exposes robustness gaps missed by standard and hand-designed user simulators while maintaining conversation realism. This is achieved through lightweight prompt optimisation alone, without model fine-tuning.

★ A Unified Arena for LLM Agents across Diverse Eurogame

Fengrui Zhang, Xiang Li, Ruomeng Ding, Jiayi Yuan, Yuchen Huang, Zhaoyang Han — Tue, 27 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Interactive environments offer a practical way to evaluate large language model (LLM) agents, and competitive games extend this evaluation to decisions shaped by competing goals and other agents' choices. Yet existing evaluations confined to individual games or execution frameworks leave unclear which competitive advantages persist across settings. We introduce DECE (Diverse Eurogame Competitive Environments), a unified arena spanning ten Eurogames with shared interfaces and evaluation protocols. These games combine lasting strategic commitments, competition over shared resources, and multiple routes to victory. Our primary study evaluates 12 model-harness configurations, spanning nine models and four execution harnesses, in 480 three-player matches across ten games. The main and supplementary studies comprise 1,032 completed matches in total. These comparisons reveal broad cross-game ranking agreement alongside game-specific strengths. Matched studies show that the same model can perform differently across execution harnesses, adding another source of variation to these profiles. These performance advantages do not consistently translate into better token or cost efficiency, making resource use an additional consideration when comparing configurations. Together, the findings support evaluating complete agent configurations by both their competitive performance across games and the resources needed to achieve it.

★ OpenESM: Energy-Steered Models Are Scalable Foundation Learners

Yuan Pu, Yudong Lu, Ting Wang, Zhicong Hong, Qinuo Li, Xuansheng Wu — Tue, 27 Oct 2026 00:00:00 +0000

96% · New model 96%, Evaluation 39%

Energy-based models have shown promising properties for domain-specific tasks but remain challenging in modeling general-purpose scenarios. To bridge this gap, we design energy-steered models (ESMs) for foundational language modeling. ESMs integrate energy steering and token generation within a single architecture to leverage the energy-guided search to refine token generation. Specifically, ESMs first optimize a latent state over the energy landscape, then leverages the refined latent state for token generation. By performing such energy-guided pre-training on large-scale corpora, ESMs learn generally effective energy landscape where favorable tokens correspond to lower energy values. To explore the scaling behavior of ESMs, we pre-train ESMs on three large-scale corpora, then evaluate across 26 downstream tasks spanning a broad range of domains. Our results demonstrate that ESMs have promising scaling behaviors with increasing model size and training data, and outperform GPTs on most downstream tasks. These findings highlight ESMs as a promising architecture for large-scale pre-training and general purpose problem solving. We fully release the model weights and training infrastructure to enable the community to completely reproduce and extend our results. GitHub: https://github.com/datamllab/openesm HuggingFace: https://huggingface.co/collections/guan-wang/openesm

★ Linear-Time Context Compression with State-Space Encoding

Tue, 27 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, New model 48%

Transformer-based large language models (LLMs) impose substantial computation and memory costs as context sizes increase, motivating methods that compress a context into a shorter representation before inference. Existing soft-prompt compression methods typically use the same Transformer backbone as used in the target model, where producing the compressed representation leads to significant latency overheads. This restricts them to the offline setting where context is compressed once and then reused across many queries. We propose the Boa Compressor, a plug-and-play online context compressor built on a Mamba state-space model (SSM) that encodes arbitrary-length context in a single linear-time pass, producing a representation on which we condition a frozen target LLM. The SSM is trained by distillation, matching the next-token logit distribution from the LLM ingesting the full context (as teacher) to the same LLM ingesting the compressed representation (as the student). At equal compression ratios, Boa matches or surpasses the performance of existing state-of-the-art offline compression baselines, while being an order of magnitude faster than Transformer-based implementations, thereby reducing end-to-end latency rather than adding compression time overhead. We demonstrate that Boa also scales well to longer contexts and to compression ratios as aggressive as 16x. By enabling effective compression that is cheap enough to run on every query at inference time, Boa becomes a practical and general tool for efficient long-context inference.

★ TraceQuery: learning what to recompute from full-denoising trajectory in diffusion language models

Yukai Jin, Zhankai Ye, Jiayue Yang, Fangyi Li, Shuoqiu Li, Xu Gao — Tue, 27 Oct 2026 00:00:00 +0000

96% · Efficiency 96%

Diffusion large language models (dLLMs) enable flexible generation through iterative masked denoising, but incur substantial inference cost because token representations are repeatedly recomputed across denoising steps. Existing cache-based methods reduce this cost by reusing intermediate results across denoising steps and recomputing only a subset of positions at each step, chosen either on a fixed schedule or by heuristic signals of change. Our trajectory experiments instead point to decoding trajectory as the more relevant quantity. Cached generation remains accurate when it stays close to the full-denoising trajectory, with errors concentrated around unmask-order deviations and reduced as longer reference prefixes are recovered. This suggests a more direct principle: limiting representation error helps, but keeping generation on the full-denoising trajectory is a more direct route to accurate decoding. We therefore formulate efficient dLLM inference as trajectory-guided querying. Rather than broadly repairing representation error, we use full-denoising trajectories to supervise which unresolved positions should receive limited fresh computation. Based on this formulation, we introduce TRACEQUERY, a lightweight sparse decoding framework that predicts a small set of near-future trajectory-relevant positions, queries only those candidates, and lets the backbone select which one to commit from their fresh predictions. TRACEQUERY delivers the strongest quality-computation trade-off among cache-based approaches, sustaining high accuracy with substantially fewer fresh token computations, and generalizes across tasks, model variants, and families. Code: https://github.com/fish1928/tracequery.git

★ SLICE-Mem: Salience-Gated Lightweight Information Curation for Efficient LLM Agent Memory

Xiaobin Chen, Dian Jin, Lin Shu, Defeng Sun, Xiaoming Tao — Tue, 27 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, Evaluation 34%

Long-term memory enables large language model (LLM) agents to maintain personalized and temporally coherent interactions, but repeated LLM inference for memory admission, extraction, and organization introduces substantial cost and latency. We present SLICE-Mem, a lightweight memory framework that combines salience-guided admission and predicate-centered information extraction. A compact classifier identifies user turns worth retaining, and a semantic-dependency-based extractor converts admitted turns into concise memory entries. This online pipeline requires no generative LLM calls, while cross-entry reconciliation is deferred to an offline consolidation stage. We evaluate SLICE-Mem on the weekly subset of Memora and LongMemEval-S using 8B and 30B Qwen backbones for memory construction and consolidation on both benchmarks. Compared with the generative-memory baseline MemOS under the Qwen3-30B configuration, SLICE-Mem's online pipeline achieves 46.60 versus 46.68 FAMA (Forgetting-Aware Memory Accuracy) on Memora Weekly and 56.8% versus 59.3% accuracy on LongMemEval-S, while reducing total LLM token consumption by 92.4% and 99.5%, respectively. These results demonstrate that SLICE-Mem can preserve useful evidence for long-term memory with substantially less reliance on LLM-based memory construction.

★ Distilling Expert Judgment at Scale

Robbie Goldfarb, Andrew B. Hall, Jillian Fisher, Kathryn Salam, Matt Wilde — Tue, 27 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

Frontier AI models are increasingly deployed on topics where the stakes are high and expert judgment is required, including healthcare, defense, finance, governance, and politics. We present a general methodology for distilling expert human judgment into automated evaluation systems that operate at scale. Our approach assembles networks of domain experts and develops techniques for extracting not just their conclusions but their reasoning processes, encoding those processes into structured editorial standards that automated judges can apply continuously. We ground our approach in a long tradition of research on the nature of expertise and the conditions under which expert judgment is reliable and more effective than other approaches, and argue that independent, expert-calibrated evaluation is necessary not only on epistemic grounds but as a matter of accountability and public trust. We validate the methodology on AI news reporting as a first application domain. On source quality, two independent domain experts achieve Krippendorff's $\alpha = 0.83/0.69$ on a held-out test set when applying the editorial standards, and our automated judge reaches $F1 = 0.81/0.97$ against expert consensus. On neutrality, experts achieve $\alpha = .72$ on the test set, and the automated judge reaches $F1 = 0.86$ against expert consensus. Additionally, a controlled comparison shows that expert-calibrated judges substantially outperform uncalibrated frontier models on every source-quality metric, demonstrating that encoding expert reasoning is the critical step that prompt engineering alone cannot replace. We present this benchmark as a proof of concept for a broader evaluation infrastructure applicable to any domain where judgment matters and expertise is scarce.

★ Seeing the Options: Typed Decisions over Visual Candidates with Reusable Option Memory

Fengxin Li — Tue, 27 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, Evaluation 95%

Typed decision-making tasks in agents, such as UI element selection, product recommendation, and chart matching, require models to choose from a set of discrete candidates. While prior work demonstrates that text-only System One models achieve four key properties including candidate-order invariance, scalability, and cache reuse via separate request-candidate encoding, extending them to multimodal settings where candidates are visual remains underexplored. We introduce Iris, a framework that accommodates image, text, or hybrid candidates while preserving the core decision module. We prove that order invariance and exact cache reuse hold for any per-candidate encoder under Iris, and derive the closed-form latency reduction. We construct a visual-candidate benchmark from Multimodal-Mind2Web (K=16, deduplicated, order-shuffled, three generalization splits). Independent encoding (Iris) achieves 78.4 / 72.1 / 74.8 percent accuracy with zero-flip rate under candidate reordering. Joint encoding reaches 90.6 / 89.5 / 91.6 percent but suffers an 8 to 10 percent answer-flip rate when shuffled. To bridge this gap, we propose a design that encodes each candidate conditioned on the request state without inter-candidate cross-attention; it preserves strict order invariance, enables forked cache sharing, and attains 91.7 / 86.5 / 90.0 percent accuracy at one-third of OneJev latency with calibrated null probabilities (AUROC 0.93 to 0.97). Reusing image candidates across requests cuts latency by 37x at K=256 and 191x at K=2048. On MetaEncoder benchmarks, Iris-4B achieves 100 percent on JevBench original and easy, outperforming 30B MetaEncoder and 27B OpenJev. Finally, we resolve a systems bottleneck on Ampere GPUs: reformulating Qwen3.5 patch-embedding Conv3d into its exact matrix-product form accelerates processing by 350x. Code, data, and checkpoints: https://github.com/lllfx/Iris.

★ Rectified Language Flow

Baiyu Su, Runlong Liao, Lizhang Chen, Qiang Liu — Tue, 27 Oct 2026 00:00:00 +0000

96% · New model 96%, Efficiency 94%, Evaluation 30%

Continuous-state flow and diffusion models offer a promising approach to fast, parallel language generation, but matching the generation quality of autoregressive models remains challenging. We present Rectified Language Flow (ReLF), a language model that generates text by transporting noise to token embeddings using rectified flow. ReLF combines several complementary design choices: fixed token embeddings constructed with Hellinger-SVD, self-conditioning guidance trained with cross-entropy, and regularized reflow for few-step generation. We further introduce singular logits, which combine an analytically computed term that drives commitment to a valid token embedding as noise vanishes with a bounded neural residual that learns contextual token preferences. With 64 function evaluations on OpenWebText, ReLF achieves lower generative perplexity than autoregressive baselines at matched diversity across all three model sizes tested, while also outperforming prior continuous flow methods in this setting. After two reflow rounds, ReLF still outperforms autoregressive baselines with 16 function evaluations while generating 1,024-token sequences 15-19 times faster at batch size 1.

★ Reversible Pseudonymisation of Documents Sent to External LLMs: an Open Bilingual Benchmark and an Engine

Vladimir Beskorovainyi — Tue, 27 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

Professionals send working documents to external large language models. A common safeguard is reversible pseudonymisation: personal data are replaced with placeholders before the request and restored locally in the answer. We present an open benchmark of five existing open tools and a regular-expression baseline on Russian and UK English, and an open engine built on the two strongest tools. The benchmark excludes each tool's own authors' dataset from its verdict, compares the engine with each tool on identical datasets for detection, reports masking precision, and tests on a held-out set of synthetic documents whose files, hashes, decision rules, run plan and stopping rule were published before any run. Two local LLMs test whether placeholders survive. All evaluation data are synthetic, and the benchmark was run by the author of the engine; code, data, predictions and raw answers are released. On external datasets no existing tool meets all predefined thresholds: in UK English the identifier and leak thresholds are not met, in Russian the identifier threshold is borderline. On the last held-out set of 1,000 documents per language, the best existing tools leave a whole value visible in 61% (Russian) and 58% (UK) of documents. On these documents, from the generator it was developed on, the engine does so in 0.3% [0.1; 0.9] and 0.2% [0.1; 0.7], leaves any part of a value visible in 0.8% and 0.2%, and meets all detection thresholds in both languages. In UK English it also restores at least 99% of its placeholders after 760 answers. In Russian it restores only 94.6% [93.6; 95.6] of name placeholders, against a 99% threshold, mainly because the inflection it borrows from pii-guard produces wrong surname forms. The engine was developed over six earlier held-out sets from the same generator; real documents were not tested.

★ ParetoRSI-Gym: Can AI Agents Improve Language Models Across Multiple Domains?

Haojin Wang, Yangzhen Wu, Zeyi Chen, Jian Xie, Jiaying Ye, Lili Meng — Tue, 27 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Agents 72%

Recursive self-improvement (RSI) requires an agentic system to design, train, and evaluate its own successor. Data-driven post-training is a central part of that loop, yet existing evaluations measure it on a single benchmark, where an agent selects from a static pool or distills a stronger model to push one score up in one run. Real post-training is multi-domain, and a gain in one domain has to be weighed against a regression in another. We propose ParetoRSI-Gym, in which an agent searches for data, curates it, trains, reads back per-domain results, and repeats, with everything but the data held fixed: the initial LLM, the five domains (math, code, finance, medicine, science), the toolset, and the SFT pipeline. Generalization benchmarks in each domain separate capability gains from overfitting to the scores the agent optimizes against. Across three agent systems, three base models, five rounds and multiple seeds, the strongest system, Claude Code with Opus-5, raises the generalization scores of its proxy-best checkpoint by 9.2% relative to the starting model (+2.3 absolute points), averaged with equal weight over three base models and five domains, with a single-domain relative gain of up to 63% (medicine on Ministral-3-3B). Further analysis shows that the harness matters as much as the model it drives, and that each agent keeps a unique data selection strategy. We see our work as a step towards evaluation of RSI systems in real-world machine learning engineering.

★ HobbyLM: A Sparse Mixture-of-Experts Model Family on a Hobby Budget

Harish Santhanalakshmi Ganesan, Hariharan M. S., Prabhurajhan R — Tue, 27 Oct 2026 00:00:00 +0000

96% · New model 96%, Efficiency 87%, Agents 36%

We describe HobbyLM, a family of small sparse Mixture-of-Experts (MoE) models trained from scratch with a few hundred to a few thousand dollars of rented GPU time. The core is a decoder-only transformer with fine-grained experts, sigmoid routing with auxiliary-loss-free load balancing, grouped-query attention with per-head QK-normalization, and aMuon/AdamW optimizer split. Each of these choices was selected with a single-variable ablation suite at 130M parameters. On this architecture we train a 500M-parameter model (169M active) on 40B FineWeb tokens and post-train it into a chat model, a masked-diffusion language model, a function-calling and computer-use agent, and a single "omni" checkpoint that reads text, images, video, sound and speech through frozen encoders and small projectors. We run these models on a laptop CPU with a from-scratch Rust inference engine. Two of the authors then adopted the architecture without architectural changes to build HobbyLM-1B (1.04B total, 305M active parameters), trained on a 100B-token curated mixture with a high-quality anneal. HobbyLM-1B reaches a 47.5 average on seven zero-shot benchmarks, and its builders also release an instruction-tuned model and an on-device macOS computer-use adapter. Finally, we describe HobbyLM-Image, a 333M-parameter flow-matching diffusion transformer that generates 1024x1024 images in a 32x-compressed latent space, together with a dependency-light CPU implementation of the full text-to-image pipeline. We report what worked, what did not, and where the small-scale ceilings lie. We also state for each component which author designed or built it.

★ BehaviorMem: Building and Evaluating User Memory from Real-World E-Commerce Behavior Streams

Sizhe Wang, Xinghao Fang, Rui Qiu, Zhenjia Ma, Dagui Chen, Kun Kuang — Tue, 27 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

User memory helps personalized systems build on their understanding of users over time. In e-commerce, users convey their preferences and needs through both explicit language and implicit interactions with products. We study memory built from logs of these interactions, examining how it can \emph{remember} concrete evidence and established interests, and use that evidence to \emph{infer} underlying needs and preferences that inform judgments beyond well-observed product domains. Existing memory methods primarily target language-based interactions, and evaluations offer limited real behavioral data and limited separation of these capabilities. This leaves open how to jointly support and separately evaluate remembering and inference from behavioral logs. We formulate the problem of continual behavioral user memory around these joint requirements and develop a dedicated evaluation framework and memory method to address them. \textbf{BehaviorMemBench} is a benchmark built from real e-commerce logs covering 1{,}194 users and 17{,}462 behavioral batches over more than three years, with automatic future-behavior labels and a task-independent memory interface to separately evaluate remembering and inference under a common protocol. \textbf{BehaviorMem} is a memory design that organizes explicit behavioral facts and inferred user characteristics across long-term and short-term scopes in a two-axis state, with evidence-grounded formation, continual maintenance, and query-independent readout. Experiments across two builders and two task readers show that BehaviorMem leads the evaluated agentic memories on remember and infer, and reveal different method rankings across the two capabilities and competitive statistical baselines. Validation on a public dataset further supports its usefulness for downstream tasks based on behavioral histories.

★ RubriX: A Fast, Fine-Tunable Non-Autoregressive Rubric Judge

Balamurugan Paramasivam (CobuildX), Sudalai Rajkumar (CobuildX) — Tue, 27 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, Evaluation 82%, New model 80%

Rubric-based evaluation asks which level of a rubric a response meets. It is usually done by prompting a large language model (LLM) as a judge, which needs a generative call per judgment and reads every level description, so its cost grows with the rubric. We present RubriX, a rubric judge whose architecture is inspired by contrastive language models (CLM): a LoRA-adapted encoder embeds the input and each level description, and a softmax over scaled cosine similarities gives a probability for every level without text generation. Trained on public rubric data and proprietary synthetic data, RubriX reaches 75.4% exact accuracy on held-out rubrics, against 57.1% for TypeSafe's Jev and 49.8% for a Qwen3-8B judge. On eight human-labelled benchmark sets it is within 6 percentage points of Jev's accuracy on most of them, while Jev keeps a larger lead on two and ranks responses best on average; against the results reported in the G-Eval paper, RubriX outperforms G-Eval with GPT-3.5 on every SummEval dimension and both QAGS subsets. With level embeddings cached, RubriX's latency stays at about 28 ms per judgment as level descriptions grow from 10 to 3,000 tokens, while Jev's (approximate network time subtracted) rises from 44 ms to 211 ms and the Qwen3-8B judge's from 32 ms to 516 ms, making RubriX 7.7x faster than Jev and 18.8x faster than the Qwen3-8B judge at 3,000 tokens per level.

★ ClinVerse: A Realistic Multi-Agent Diagnostic Environment for Training and Evaluating Medical Agents

Sirry Chen, Fangke Chen, Xuanwen Ding, Wei Chen, Zhongyu Wei — Tue, 27 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, New model 32%

Disease diagnosis is a key application of large language model (LLM) agents in healthcare. However, existing medical agents are primarily trained and evaluated with static question-answering data, providing limited opportunities to learn active information acquisition, examination decision-making, and dynamic clinical interaction. To address this limitation, we present ClinVerse, an interactive, real-data-grounded multi-agent environment designed for training and evaluating medical diagnostic agents. Built on ClinPatient, a scalable dataset comprising 150k diagnostic cases derived from real-world electronic health records, ClinVerse supports coordinated interactions among six specialized clinical roles. Specifically, a doctor agent starts from an initial patient complaint, progressively gathers clinical evidence through patient interviews and examination requests, and ultimately produces a diagnosis. For each interaction trajectory, a rule-based scoring framework quantifies information acquisition, examination ordering, and diagnostic accuracy. These trajectory-level scores serve as reward signals for agent training while also enabling scalable evaluation. In experiments, we evaluate various general-purpose and medical-domain models and show that nine self-evolution methods across three categories consistently improve agents' diagnostic capabilities within ClinVerse. We further analyze the interplay among model capability, environment feedback, and self-evolution methods.

★ From Questions to Queries to Insights: A Schema-Grounded Agentic Workflow for Natural-Language Crash Data Analysis

Jingwen Zhu, Keshu Wu, Steven T. Parker, Bin Ran, David A. Noyce — Tue, 27 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Hallucination 67%, Agents 50%

Transportation agencies maintain extensive crash databases to identify safety priorities, evaluate countermeasures, and guide resource allocation. Making these data accessible to practitioners requires flexible ways to translate analytical questions into reliable results. Existing dashboards support predefined analyses, while additional requests often depend on specialists who understand database queries, coded fields, and agency conventions. This study presents CrashDB-QA, a fully local, schema-grounded agentic workflow that enables natural-language analysis of crash databases. Specialized components interpret questions, generate and validate read-only SQL, execute queries, and explain the results. A shared knowledge base grounds query construction in the agency's data dictionary and safety-analysis conventions, while recorded intermediate outputs support inspection and auditing. An optional extension, CrashDB-QA+, checks the alignment among the question, query, and execution result and supplies evidence for iterative repair. We evaluate the workflow on a frozen snapshot of 4.4 million Wisconsin crash records using a new benchmark of 300 questions across ten analytical categories and nine local open-weight models. Domain grounding substantially improves query accuracy, and a code-specialized 7B model provides the best accuracy-latency balance among the tested models. Compositional analyses remain more difficult than routine retrieval, particularly comparisons between groups or periods. Verification improves performance when the Judge uses authoritative field definitions and the Refiner receives execution evidence, whereas additional checking alone offers no consistent benefit. These findings demonstrate the potential of locally deployed, domain-grounded workflows to broaden access to crash analysis and identify the reasoning and verification capabilities needed for reliable use.

★ Useful Diversity or Hallucinated Variety? Persona-Conditioned Synthetic Query Generation for E-Commerce Search

Tue, 27 Oct 2026 00:00:00 +0000

96% · Hallucination 96%, Evaluation 83%

Synthetic supervision can bootstrap semantic parsers in new search domains, but prompt-induced diversity is useful only when generations are faithful to the structured intent. We explore this for cold-start e-commerce search: 118,098 synthetic pairs produced by 6 LLMs with 9 persona-conditioned prompt families on a 47-field UK automotive schema, against 1,214 human-annotated and 15,304 production queries. We show that the IPIP personality bundle improves parser F1 in all four matched contrasts (Holm-corrected $p<.001$), and all three bundles have a positive main effect over the complete $2^3$ grid. Syntax conditioning is the axis that moves generations towards real traffic. It also induces lexical diversity. Alignment sKL, the divergence between von Mises-Fisher distributions fitted to synthetic and human-annotated queries, is an unlabelled proxy for parser F1 ($\rho=-0.917$ against human queries written for simulated shopper scenarios, lower is closer). An LLM committee, checked against a blinded author pass, triages the 927 queries missed by every parser and labels 37% as hallucinated. Dropped constraints dominate. The four syntax-bearing families have the largest hallucinated shares, 38 to 42%, while IPIP alone has the lowest at 28%.

★ The Lay of the Land in Idiom-Translation Quality Estimation

Malik Marmonier, Beno\^it Sagot, Rachel Bawden — Tue, 27 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

Idioms have long been a challenge for machine translation, and while large language models (LLMs) have proven increasingly adept at handling them, whether quality estimation (QE) systems have kept pace and can reliably score idiomatic renderings of these expressions above faulty literal ones remains unclear. We introduce the first parallel dataset pairing idiom-bearing English sentences with both literal and idiomatic translations across nine target languages, augmented into multi-sentence segments and a graded literality continuum. Using this dataset, we benchmark eleven conventional QE metrics from the COMET and MetricX families and ten LLMs from the Qwen-3, Gemma-4 and OLMo-3 families used as GEMBA-style judges. We find that most systems, especially older metrics and smaller models, display a marked literality bias, rating literal translations of idioms above idiomatic ones. Higher parameter counts and newer training sets appear to reduce this bias to an extent; nevertheless, of all the benchmarked systems, only MetricX-24-Hybrid (XL, XXL) and Gemma-4-31B consistently rate idiomatic translations of idioms above literal ones, and only Gemma-4-31B reliably ranks intermediate degrees of idiomaticity and literality. Our findings caution against deploying standard QE metrics in domains rich in figurative language, where higher scores might be rewarding undesirable literalism.

★ MANTIS: Multi-Axis News Transformations for Instance-Wise Stress Testing and Worst-Case Training in Fake News Detection

Dexin Liu, Fan Yang, Zimu Wang, Xunfei Zhu, Bolin Ren, Bingyan Zhang — Tue, 27 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, New model 88%

Automated fake news detection is vital on social media, where fake news is easily rewritten to fool detectors. However, existing approaches typically rely on superficial writing cues or single-style text rewrites, resulting in poor robustness under diverse cosmetic modifications. We construct MANTIS-Bench, a multi-axis presentation-attack benchmark that pairs original news with rewritten views across writing style, emotional expression, and dissemination framing. We further propose MANTIS, an instance-adaptive adversarial framework enabling dynamic worst-case training tailored to each news article with zero inference-time overhead. This reframes fake news defense from heuristic format smoothing into targeted, fact-invariant credibility assessment. Experiments on MANTIS-Bench show that MANTIS reduces the Union Attack Success Rate (UASR) by 3.55 percentage points relative to SheepDog on GossipCop and consistently outperforms prior single-axis defenses on Macro-F1 while maintaining competitive clean accuracy, demonstrating the effectiveness of multi-axis and instance-adaptive worst-case training.

★ CANS: Capability-Preserving Neuron Steering for Large Language Models

Yucheng Wu, Chengyi Xie, Chenpeng Wang, Liangming Pan — Tue, 27 Oct 2026 00:00:00 +0000

96% · Hallucination 96%, Evaluation 59%, Efficiency 32%

Activation steering offers training-free and dynamic control of target behaviors in large language models, but existing methods can impair general capabilities and require domain-specific intervention designs. We introduce Capability-Preserving Neuron Steering (CANS), a unified framework that balances target-behavior control with capability preservation. To make target-behavior improvement and general-capability preservation tractable, we introduce two proxies: log-probabilities of target tokens in model generated responses, and output-distribution KL divergence on general-purpose inputs. Combining the local approximation of the two proxies yields a closed-form solution, enabling lightweight steering without iterative model training or additional inference-time computation. The resulting neuron-wise scaling coefficients achieve effective steering while limiting general-capability loss. A shared data-construction and intervention procedure supports adaptation across domains using sampled responses with target-behavior filtering and target-span annotations. Across safety, toxicity, and hallucination control on Qwen and Llama models, CANS achieves comparable or better steering while better preserving MMLU, BBH, and IFEval performance than SFT, CAA, and neuron-level baselines. On Qwen3-4B, CANS reduces HarmBench attack success from $15.50\%$ to $1.75\%$ with a $1.11$-point decrease in average capability. These results support CANS as a practical approach to behavioral control across domains.

★ EXCISE: Locating and Excising Distracting Sentences in Medical Question Answering

Yerim Hong, Sohyun Son, Heeseung Chung, Mujeen Sung — Tue, 27 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

Language models are increasingly considered for clinical decision support, where their answers should be grounded in clinical evidence and remain robust to irrelevant information. Prior work shows that a single irrelevant sentence can lower the accuracy of medical models, but not where inside the model its effect can be removed, or how to do so without knowing which sentence it is. We find that blocking attention to a known distractor restores the answer, whereas head masking can suppress correct answers. We also show that the model's hidden states mark the distractor as the sentence that least fits the case even when the model follows it. To remove a distractor's effect without knowing which sentence it is, we propose EXCISE, an inference-time method that blocks the sentence most influencing the answer, but only if the model's hidden states mark it as inconsistent with the case. On MedQA-TD (MedQA with Targeted Distractors), a benchmark we construct in which each distractor targets a specific wrong option, EXCISE gains 16.7-22.8 accuracy points at a 1.5-2.1-point cost on the same questions without distractors, outperforming prompting, rewriting, and steering baselines for all three models. Its gains extend to MedDistractQA and BiasMedQA, as well as chain-of-thought generation.

★ Quantifying Claim Smoothing by LLM Research Agents

Tiasa Singha Roy, Manya Wadhwa, Greg Durrett — Tue, 27 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Hallucination 73%, Agents 39%

Deep research, or any use of large language models (LLMs) to read a large document collection and produce insights, requires synthesizing information from different sources. In this work, we find that LLMs produce long-form research reports that read as confident syntheses but which fail to accurately reflect the underlying documents. We call this failure smoothing. In collapsing diverse and potentially conflicting sources into a single narrative, the model discards the evidential "jaggedness" of its corpus and asserts conclusions more decisive than the citations can support. We introduce a claim level scoring pipeline, SmoothScore, to quantify this effect. Our pipeline extracts key conclusive claims from model generated reports, then finds evidence relevant to each claim and clusters that evidence along high level aspects. Finally, each claim is scored with respect to the evidence to quantify the degree of smoothing done by the agent. Across 7 systems and 4 datasets, we find that agents consistently produce stronger claims than warranted.

★ Learning When to Commit: Schedule Distillation for Fast Diffusion Language Models

Aidar Myrzakhan, Zhiqiang Shen — Tue, 27 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, New model 92%, Evaluation 33%

Diffusion large language models (dLLMs) can generate multiple tokens in parallel, but often require many decoding rounds to complete a response. Efficient generation depends not only on which tokens are easy to predict, but also on how committing to them helps predict the remaining positions. A highly predictable token may provide little useful context, while revealing a less predictable token can help several other positions become predictable sooner. We propose dSpark, a new schedule distillation-based framework for learning when to commit tokens for faster diffusion language generation. To achieve this, we construct unmasking schedules that prioritize tokens whose reveal makes other positions easier to predict, reducing uncertainty earlier in decoding. We then distill these schedules into the model to enable faster parallel sampling, with a KL-divergence regularizer to help preserve generation quality. We extensively evaluate our approach across mathematical reasoning and code generation tasks. On GSM8K, our method achieves remarkable 10.12$\times$ and 5.12$\times$ speedups over the various baselines on LLaDA-8B-Instruct and Dream-7B-Instruct, respectively, while maintaining superior task performance. Our code is available at https://github.com/VILA-Lab/dSpark.

★ StreamVibe: Proactive Video Agents for Evolving User Intent

Chengzhi Liu, Yuxiang Lai, Yuzhe Yang, Yanqing Liu, Jiheng Hou, Yue Cao — Tue, 27 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, New model 62%, Efficiency 30%

Vision-language models are shifting from offline passive understanding toward online proactive interaction. However, prevailing methods largely rely on explicit requests while miss the implicit user intent (e.g., "where should I click next?" while a user navigate a website), meanwhile fail to update the state simultaneously as user requests might update continuously. These limitations make them less suited to intent-aware interaction in practical. To address these challenges, we introduce StreamVibe-Bench, a new benchmark serving as testbed for user intent interaction, with 4.1K evaluation samples and 85K training samples spanning explicit, implicit, and dynamic intents. Building on this benchmark, we develop StreamVibe-VL, a streaming proactive model that infers implicit user intents from continuous observations and dynamically tailors its interactions as these intents evolve. We further introduce StreamVibe-Harness, which maintains user intents as persistent states with a fleixble lifecycles via a hierarchical memory to track the latest user intent and environmental state. Our experiments reveal three key findings: (i) StreamVibe-VL preserves strong general video understanding while improving response timing and answer quality by 10.7% over other baselines; (ii) existing proactive models remain limited in understanding user intent and deciding when to respond, especially under implicit and dynamic intents; and (iii) StreamVibe-Harness further improves response consistency and adaptation over sustained interaction, resulting in a 12.3% gain in proactive interaction.

★ EviGround-CT: Sentence-to-Volume Grounding for Verifying LLM-Generated Chest CT Reports

Ziyuan Li, Juncheng Ni, Feng Li — Tue, 27 Oct 2026 00:00:00 +0000

95% · Hallucination 95%, Evaluation 86%

Chest computed tomography (CT) reports generated by large language models (LLMs) can describe positive findings that lack localized evidence in the scan. We propose EviGround-CT to verify these findings after report generation. A deterministic extractor identifies positive finding sentences, and a 3D grounding model predicts an evidence mask for each sentence. Using the full sentence retains anatomical location, laterality, and descriptive modifiers in the grounding query. We evaluate report generation on CT-RATE, and grounding and verification on a held-out 100-case ReXGroundingCT subset. Using the available annotations, sentence-grounded verification raises evidence-supported precision (ESP) from 0.263 to 0.365 across 137 extracted in-scope positives, compared with category conditioning. The corresponding unsupported positive finding rate falls from 0.737 to 0.635. The paired case-level bootstrap ESP difference is +0.102 (95% confidence interval: [0.028, 0.181]). The predicted masks associate individual report statements with spatial evidence that can be inspected alongside the CT.

★ One Call Ahead: Packed Self-Speculative Decoding with Diffusion Drafting

Yuxiao Chen, Changxiao Cai — Tue, 27 Oct 2026 00:00:00 +0000

95% · Efficiency 95%

Autoregressive (AR) language models generate tokens sequentially, so generating long outputs is costly. Speculative decoding mitigates this by using a lightweight drafter to propose multiple tokens for the target AR model to verify in parallel, while a diffusion-based drafter can produce an entire draft block in one forward pass. In models with AR and diffusion modes, one network both drafts and verifies, but drafting still waits for verification, so the standard schedule needs two sequential model calls per iteration. Existing one-call schedules avoid the wait but incur input size quadratic in the speculative width. Motivated by this efficiency bottleneck, we introduce Packed-SS, a self-speculative decoding schedule that verifies and drafts in a single model call with input size linear in the speculative width, while preserving the target model's distribution. The key idea is to let the diffusion drafter run one call ahead: while verifying the current draft, the model simultaneously constructs the next draft from an older verified prefix and keeps its unresolved suffix for the next iteration. Under a mild monotonicity assumption on the drafter, we further show that Packed-SS requires no more model calls than the standard two-call decoding in the greedy setting. Experiments on Nemotron 3B/8B and SetDLM 1.7B show higher throughput and lower mean wall time than both baselines under greedy decoding.

★ SkillShift-Bench: Benchmarking Skill Learning under Environment Evolution for Web Agents

Bo-Yu Chen, Zhi Zhou, Yu-Feng Li — Tue, 27 Oct 2026 00:00:00 +0000

95% · Agents 95%, Evaluation 94%

Skill learning enables web agents to reuse prior experience for long-horizon interactive tasks. However, learned skills are typically evaluated in closed environments, overlooking the fact that real-world web deployment involves continuous shifts in invocation context, perceptual grounding, and execution feasibility. To this end, we introduce SkillShift-Bench, the first benchmark for evaluating skill learning with controlled open-environment shifts. SkillShift-Bench simulates Contextual Shift, Perceptual Shift, and Execution Shift across three self-hosted web systems, and evaluates agents with diagnostics covering task success, interaction cost, skill use, and skill failure. Comprehensive experiments reveal systematic robustness degradation under controlled shifts, characterized by lower success rates, higher interaction costs, and more frequent skill failures, with Execution Shift exerting the strongest pressure. We further formulate shifted skill execution as a nonstationary local executability problem and provide both a dynamic-regret analysis and a practical test-time adaptation method, demonstrating the feasibility of improving learned-skill execution without relearning the skill library. Code, tasks, environments, documentation, and the leaderboard are available at https://skillshift-bench.github.io/.

★ Beyond Prompts: A Systematic Study of LLM-Based Machine Translation Evaluators

Tue, 27 Oct 2026 00:00:00 +0000

95% · Evaluation 95%, Efficiency 80%

LLM-based evaluators have rapidly become a standard tool for machine translation evaluation, yet the design choices that shape their behavior remain poorly understood. We present a systematic study of how instruction design, few-shot demonstrations, backbone model choice, and multi-step workflows affect span-level and ranking-based MT evaluation. We collect new expert RATE annotations on WMT24 and WMT25 English-Russian translations, with transfer validation on MQM- and ESA-annotated benchmarks across five additional language pairs. Our experiments reveal four robust findings: instruction structure shapes evaluator behavior, high-quality demonstrations matter more than quantity, frontier backbones exhibit distinct evaluator profiles (recall-, precision-, or ranking-oriented), and fusing outputs across diverse backbones outperforms single-evaluator multi-call setups. Although evaluator design exposes a large combinatorial space, several discovered design principles transfer consistently across annotation protocols and language pairs. We synthesize these findings into ORBIT, a family of LLM-based MT evaluators. Our strongest configurations achieve expert-level performance in error span detection and in Accuracy and Fluency ranking at roughly 20x lower cost than expert human annotation.

★ Refined Assessment for Translation Evaluation: Rethinking Machine Translation Evaluation in the Era of Human-Level Systems

Tue, 27 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

As machine translation systems approach human-level quality, traditional evaluation methodologies struggle to detect subtle translation errors. We critically examine limitations in current gold-standard approaches (MQM and ESA), including inconsistencies from variable annotator expertise, excessive categorization complexity, coarse severity granularity, accuracy bias over fluency, and time constraints. To address this issue, we introduce a high-quality dataset consisting of human evaluations for English-Russian translations from WMT24, created by professional linguists. We show that expert assessments without time pressure yield substantially different results from standard evaluations. To enable consistent and rich annotation by these experts, we developed the RATE (Refined Assessment for Translation Evaluation) protocol. RATE provides a streamlined error taxonomy, expanded severity ratings, and multidimensional scoring balancing accuracy and fluency, facilitating deeper analysis of MT outputs. Our analysis, powered by this expert dataset, reveals that state-of-the-art MT systems may have surpassed human translations in accuracy while still lagging in fluency - a critical distinction obscured by existing accuracy-biased metrics. Our findings highlight that advancing MT evaluation requires not only better protocols but crucially, high-quality annotations from skilled linguists.

★ EdgeDeID: A 0.6B Clinical De-identifier Trained Without Real Patient Data

Sina Daneshgar, Raya Namazi, Amirali Khosravi, Pouria Rouzrokh, Bardia Khosravi — Tue, 27 Oct 2026 00:00:00 +0000

95% · New model 95%, Evaluation 83%, Efficiency 53%

De-identifying clinical notes with a hosted language model discloses the information the step is meant to protect. We present EdgeDeID, a 0.6B-parameter decoder-only de-identifier that runs locally and is trained without real patient text: three LLM teachers write synthetic notes, a single labeler re-annotates them into one convention, and the model learns to emit a typed XML schema of the identifiers it finds. On a 100-note synthetic in-domain benchmark with manually verified annotations it reaches 0.994 micro-F1 and leaves no missed identifier in 99% of notes, against 0.857 and 12% for the strongest of eight off-the-shelf systems. Quantized, it runs on CPU in 615 MB. Of three additions to supervised training, only a verifier adapter helps: it separates complete from incomplete outputs of other systems at 0.94 AUROC, whereas reinforcement learning with a recall-weighted reward and a rule-based post-processing layer leave in-domain results unchanged. On MIMIC ICU nursing notes F1 falls to 0.48. The errors follow two properties our synthetic notes lack, telegraphic style and low identifier density, and three hand-written in-context examples recover 0.08 F1 without annotated target data.

★ Beyond Belief: Evaluating and Mitigating LLM Susceptibility to Misinformation

Emran Kanaan, Jiho Lee, Ibrahim Odat, Soroush Vosoughi, Weicheng Ma — Tue, 27 Oct 2026 00:00:00 +0000

95% · Agents 95%, Evaluation 91%

Tool-using large language models (LLMs) must decide whether external information warrants action. We investigate whether their expressed credibility judgments reliably indicate behavioral resistance to misinformation. We evaluate five LLMs on 196 constructed false notices paired with legitimate counterparts and 204 archived attacks adapted to simulated tool use. When asked to act on the constructed notices, models issue harmful tool calls in 27%--99% of cases they have rated untrustworthy. Removing the preliminary rating step increases harmful execution by 15 percentage points on average, showing that credibility elicitation itself changes the behavior being evaluated. Interventions that improve credibility discrimination do not consistently yield safer actions. A simple forewarning significantly reduces harmful execution in four of five models, but neither tested generated-example prebunk outperforms it in the main evaluation. These reductions can also compromise legitimate task completion. Supervised fine-tuning on generated false notices reduces harmful execution on held-out archived attacks under the rating-first protocol, with model-dependent utility costs and weaker protection without rating elicitation. Our findings motivate evaluating misinformation defenses through both harmful actions and legitimate task completion, while accounting for the behavioral effects of the evaluation protocol.

★ QECGBERT: A Scalable ECG Foundation Model Based on Self-Supervised Language Learning via VQ-VAE Waveform Tokenization

Aruna Mohan, Burcu Ozek, Or Zilbershot, Fatemeh Afghah — Tue, 27 Oct 2026 00:00:00 +0000

95% · New model 95%, Efficiency 82%, Evaluation 40%

AI foundation models play a pivotal role in advancing heart disease diagnosis from electrocardiogram (ECG) data, particularly for rare diseases where the lack of sufficient labeled data hinders supervised learning. However, most self-supervised learning methods treat ECGs as generic sequence or image data without exploiting underlying physiological structure. To address this limitation, we propose QECGBERT, a self-supervised ECG language model that explicitly encodes physiological semantics by tokenizing the P, QRS, T, and isoelectric background waveforms. Vector-quantized variational autoencoder (VQ-VAE) tokenization is applied to each 12-lead ECG, with tokens enriched by continuous waveform features to preserve morphology lost during quantization without increasing sequence length. This design enables compact BERT masked language modeling with short sequences, improving scalability and reducing computational cost relative to long-sequence approaches. QECGBERT applies BERT pre-training to the MIMIC-IV dataset containing over 800,000 unlabeled ECG recordings, and fine-tuning on labeled PTB-XL, CPSC2018, and Chapman--Shaoxing--Ningbo datasets for evaluation. The results demonstrate that QECGBERT achieves comparable or superior performance to published benchmarks, offering a scalable, physiologically aware approach to ECG foundation modeling for cardiac diagnosis.

★ Steering the Teacher: Improving On-Policy Self-Distillation for Thinking Models

Tiejin Chen, Xiaoou Liu, Hanyang Chen, Hua Wei — Tue, 27 Oct 2026 00:00:00 +0000

95% · New model 95%, Hallucination 95%

On-policy self-distillation (OPSD) trains a student on its own reasoning trajectories using token-level supervision from a reference-conditioned teacher initialized from the same model. However, access to the reference can induce teacher behaviors that undermine independent reasoning in thinking models, leading to reference hallucinations, where the student cites an unavailable reference to justify reasoning steps, and reconsideration suppression, where the teacher discourages verification and revision at uncertain steps. We introduce Steered-Teacher OPSD (ST-OPSD) to jointly address these failure modes through activation steering. ST-OPSD extracts two directions from model-generated reasoning trajectories and applies them to the teacher's hidden states during distillation target computation, suppressing reference-invoking behavior and encouraging reconsideration with marker-family calibration. The student learns from the resulting teacher distributions without being steered. Experiments on AIME and HMMT show that ST-OPSD improves accuracy over vanilla OPSD and other interventions while restoring reconsideration and reference-hallucination rates to base-model levels.

★ T2D: Target-to-Draft Model Pruning & Adaptation for Efficient Speculative Decoding

Shahriar Kabir Nahin, Xiao Ye, Muhao Chen, Ben Zhou, Anshuman Chhabra — Tue, 27 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, New model 74%, Evaluation 46%

Speculative decoding significantly accelerates autoregressive text generation in Large Language Models (LLMs), but its efficacy depends largely on two key factors: the distributional alignment between the draft and the target model, and the draft model's latency during inference. However, existing approaches typically utilize independently pretrained small draft models or seek to prune the target model to a manually chosen architecture, thereby failing to explicitly balance the trade-off between draft model alignment and token generation efficiency. To address this gap, we formulate draft model design as an optimization problem and propose T2D, a framework that constructs draft models directly from the target model using latency-aware structured pruning and knowledge distillation. In this manner, T2D obtains draft models that are jointly optimized for both measured inference latency and cross-entropy loss on the target model's responses, resulting in superlative speculative decoding performance. Through extensive experiments on multi-domain speculative decoding benchmarks and across several target model families of differing scale, we demonstrate how T2D prunes up to 92\% of the target model to obtain highly performant draft models compared to competitive baselines. Moreover, draft models obtained via T2D can also be utilized by other speculative decoding token-alignment and knowledge distillation methods in a plug-and-play fashion, leading to further gains. Our findings underscore the importance of jointly navigating the draft model architecture-alignment frontier to unlock the full potential of speculative decoding for LLMs.

★ ThinkBit: Search Codewords, Recalibrate Scales, Select by Reasoning

Rhea Sanjay Sukthanker, Pashmina Cameron — Tue, 27 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, New model 89%, Evaluation 60%

Reasoning language models generate thousands of tokens per answer, and generation speed is limited by memory bandwidth, so storing weights in two bits instead of bfloat16 can potentially make inference 8$\times$ cheaper. Two-bit methods are usually judged by perplexity and multiple-choice question-answering accuracy, both of which say little about whether a model can still reason through more difficult problems. We present ThinkBit, a strictly scalar two-bit framework built to preserve reasoning, in which every weight is an integer codeword times a bfloat16 scale shared by a group of weights. We first use a sensitivity metric to assign each layer a ternary, 2-bit, or 3-bit tier to allocate bitwidths optimally among layers. This is followed by a layerwise Gumbel-softmax-based scalar quantization algorithm that searches for optimal codewords block by block. Since a layerwise scheme does not capture how quantized layers interact with each other, we recalibrate scales, with the codewords frozen, using a knowledge-distillation-based loss against the teacher model. We then perform a short quantization-aware training (QAT) stage that lets the model mimic the base model's distribution by reassigning codewords. Finally, a paired reasoning test on training problems selects the final checkpoint between the recalibration stage and the QAT stage. On Qwen3-1.7B at matched size, our recalibrated checkpoint scores 46.7 on MATH-500 against 30.0 for EfficientQAT trained on the same data, and 73.0 after QAT on reasoning traces; at 0.52 GB it is smaller than every published 2-bit GGUF of the model yet more accurate on MATH-500. We show that this framework scales and up to 14B, a 3.74 GB model scores 94.8 on MATH-500, against 97.5 for the 29.5 GB bfloat16 model.

★ MechAlign: Evaluating Semantic Alignment at the Project Level in Mathematical Formalization

Hao Shen, Ruichen Qiu, Lihong Zhi, Xiao-Shan Gao — Tue, 27 Oct 2026 00:00:00 +0000

95% · Evaluation 95%, Agents 34%

Lean verifies that a formal proof establishes its stated proposition under declared assumptions, but does not check whether that proposition faithfully expresses the mathematical claim in a manuscript. Assessing this correspondence requires interpreting the main theorem through its project dependencies, which may be scattered throughout a large development. We introduce MechAlign, a framework for project-level semantic alignment review that combines formal dependency extraction with LLM-based assessment. Its extraction tool, LeanTrace, constructs a statement dependency graph from elaborated declarations, retaining definition bodies and structure constraints while excluding proof bodies and stopping at an explicit library boundary. The reviewer uses this context to investigate alignment and report discrepancies with localized evidence. We conduct a preliminary review of 201 main results from 173 OpenAI Math projects, using supplied informal descriptions and available paper context. A hyperbolic-group case illustrates how dependency inspection localizes a missing dimensional constraint relative to the manuscript, consistent with the upstream declaration of formalization scope. We also construct LeanReviewBench, comprising 501 fixed claim-statement-context triples from OpenAI Math, Fermat's Last Theorem, and Kakeya, with description provenance recorded. The resource supports controlled comparisons of more project-level semantic alignment evaluations.

★ EvoRouter: Cost-Efficient Multi-Agent Routing via Insight-Guided On-Policy Distillation

Yitao Xiao, Xu Zhang, Ye Cheng, Shaoyong Guo, Yan Sun, Ziyue Zhu — Tue, 27 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, New model 86%, Evaluation 38%

Serving web queries at scale with multi-agent systems requires jointly configuring large language models, collaboration modes, agent counts, interaction rounds, and skills. However, the resulting configuration space grows combinatorially, while sparse execution rewards provide limited fine-grained supervision for interdependent planning decisions. We propose EvoRouter, a self-evolving multi-agent routing framework driven by insight-guided on-policy distillation. A dedicated expert periodically synthesizes performance-cost feedback from configuration trajectories across tasks into reusable conditional routing insights about capability requirements, returns from resource scaling, and interactions among configuration choices. A privileged teacher combines these evolving insights with current execution feedback to provide dense token-level supervision for the router's own planning trajectories through on-policy distillation. Cost-aware reinforcement learning jointly optimizes task performance and computational cost. Together, these training signals internalize execution insights into the routing policy. The updated policy generates new trajectories that support subsequent insight extraction, forming an execution-driven self-improvement loop. We evaluate EvoRouter on 19 in-domain datasets spanning five domains and two out-of-domain benchmarks. EvoRouter achieves an average in-domain accuracy of 68.69%, outperforming the strongest routing baseline by 7.23 percentage points while reducing inference compute cost by 50.95%. Our code is open-sourced at https://github.com/CenYangtze/EvoRouter.

Conceptizer: Next-Token Prediction Is Not All You Need

Haoran Zheng, Chen Shani — Tue, 27 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 47%

Next-token prediction (NTP) rewards only the token that occurs next, penalizing every alternative. But natural language often offers many ways to express the same meaning. We introduce Conceptizer, a training framework that extends next-token prediction with a reward for meaning. Conceptizer uses a fixed, model-agnostic vocabulary of ~9K concepts, built from contextual synonyms, as units of meaning alongside tokens as units of text. It maps each next-token distribution to a concept distribution and shares the reward for the target meaning across its possible wordings. This provides a concept-level training signal without adding model parameters. Applied to 3 model families at 1B-8B parameters, Conceptizer achieves the lowest held-out NTP perplexity at concept-target positions and assigns the highest probability to contextual synonyms, against all 3 baselines. It also improves downstream performance on knowledge and sentiment tasks while remaining comparable on commonsense. These results show that language models can be directly optimized to predict meaning, not just the next token.

Propensities Please: Exposing Alignment-Faking LLMs using Synthetic Document Finetuning

Ardy Haroen, Florian Lorkowski, Samuel Nellessen, Felix Hofst\"atter — Tue, 27 Oct 2026 00:00:00 +0000

94% · Evaluation 94%, New model 80%

Alignment evaluations aim to assess whether large language models would engage in dangerous misaligned behaviour. Misaligned and evaluation-aware models might engage in alignment faking, where they pursue a hidden goal, but conceal this behaviour when they suspect effective oversight such as during training, evaluation or audited deployment. Such a model might view any prompt-based intervention that aims to reveal latent propensities as part of the evaluation, calling into question the reliability of existing alignment evaluations. Current methods for detecting alignment faking rely on chain-of-thought monitorability, highly realistic environments that reduce evaluation awareness, or mechanistic interventions, all of which are expected to become less reliable for future model generations. We present a method for revealing latent misaligned propensities that does not rely on any of these. We use synthetic document finetuning to teach models that misaligned actions will go undetected whenever certain keywords appear in the prompt. We place alignment-faking model organisms in environments that are clearly identifiable as honeypots and measure how their behaviour changes after finetuning. We observe that finetuning consistently widens the gap in misaligned action rates between prompts with and without the cue, indicating that finetuning allows exposing latent propensities more reliably than purely prompt-based methods. We argue that behavioural methods for detecting alignment faking and related behaviours are critical going forward, as models become more capable and less interpretable. We find that SDF often damages the models' reasoning capabilities and introduce a mitigation based on on-policy distillation.

Benchmarking System One Models for Time-Series Decision Making: A Case Study of Jev

Jianye Wang, Kexin Zhang, Jiayin Xiong, Hao Xue — Tue, 27 Oct 2026 00:00:00 +0000

94% · Evaluation 94%

Time-series tasks require both accurate predictions and fast decisions. Advances in large models have introduced decision-oriented models that directly output answers and their probabilities, which we call Structured System 1 models. Their applicability to time-series tasks remains underexplored. Using Jev 1.13.0 as a case study, we evaluate time-series classification, event type prediction, oil temperature direction and stock movement prediction, converting continuous forecasting targets into discrete decisions to fit the model interface. We establish a unified evaluation framework using accuracy and AUC. Results show stronger performance on selected explicit temporal patterns and event type prediction, but near-chance discrimination on many numerical sequence classification tasks and stock movement prediction. Some datasets also reveal a gap between ranking quality and final decision accuracy. These findings motivate improvements in temporal representation, domain adaptation, and decision-score alignment. Standardized sample construction, model calls, and result recording support extending the evaluation to future models.

More: Mode-Optimized Reasoning with Self-Adaptive Effort for Self-Evolving Agents

Dianxing Shi, Dingjie Fu, Tianyu Lei, Wangjie Qiu — Tue, 27 Oct 2026 00:00:00 +0000

94% · New model 94%, Efficiency 90%, Evaluation 33%

Self-evolving training improves the performance of agents on difficult tasks through verifiable feedback, but it often applies costly deliberation uniformly across tasks. To address this problem, we propose Mode-Optimized Reasoning with Self-Adaptive Effort (MORE), a self-evolving training framework that adaptively selects both reasoning form and computational effort for each task. Specifically, MORE follows a two-stage training paradigm: (i) \emph{Stage 1} establishes a compact prior over direct answering, short and long chain-of-thought, and code-based reasoning, and (ii) \emph{Stage 2} self-evolves this prior with Mode-Adaptive and Test-time Effort PPO (MATE-PPO) on generated, verifiable task streams without curated data. By learning mode selection and token expenditure jointly, MORE allocates concise reasoning to tasks that permit it while retaining extended deliberation where it is useful. Across coding, mathematics, and general-reasoning benchmarks, MORE achieves consistent performance gains on Base and Coder backbones at model scales from 3B to 14B. During inference, MORE demonstrates higher accuracy and better efficiency, with markedly shorter response lengths and reduced inference time.

TriageRAG: Failure-Aware Self-Training for Adapting Retrieval-Augmented Language Models to Specialized Domains

Samarth Varshney, Ritu Yadav — Tue, 27 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 53%, Hallucination 45%

Retrieval-augmented generation (RAG) lets large language models (LLMs) answer knowledge-intensive questions from retrieved evidence, but adapting a general-purpose RAG system to a specialized domain such as medicine is difficult when only an unlabeled corpus is available. Self-training approaches such as SimRAG address this by letting the LLM write its own question-answer pairs from domain passages and fine-tuning on the pairs that survive a round-trip consistency filter. We observe that these approaches treat every retained example identically: they never ask whether the student already knows the answer, already extracts it correctly from the retrieved context, or fails - and if so, whether the failure lies with the reader or with the retriever. To tackle this, we propose TriageRAG, a failure-aware self-training approach that probes the student on each self-generated question with and without retrieved context, triages it into one of four categories (known, solved, context-failure, retrieval-failure), and routes each category to the training signal that addresses it, including a parametric route that teaches the reader to answer when retrieval misses. Probing is cheap and repeated every round, so the training mixture follows the student, and the triage histogram doubles as a label-free diagnostic of where the domain gap lies. Experiments on 11 datasets spanning three domains with a Qwen3-4B backbone, and on CS-Bench with Qwen3-8B, demonstrate that TriageRAG outperforms a budget-matched SimRAG by 1.2-6.4 points and is the only method that recovers the closed-book level in every case where retrieval hurts the base model. A single ablation identifies the mechanism: round-trip filtering removes every training context that lacks the evidence, and a reader that never sees such contexts learns to over-trust noisy retrieval.

Build, Don't Play: Self-Improving LLM Research Agents Develop a Hearthstone Battlegrounds AI That Rivals Top Players

Changlian Ma, Limin Wang — Tue, 27 Oct 2026 00:00:00 +0000

94% · Evaluation 94%, Efficiency 48%, Agents 34%

Can LLM agents carry a hard real-world problem from raw data to a system that holds its own against people? We study this in Hearthstone Battlegrounds, an eight-player auto-battler with stochastic shops and battles, hidden information and nearly a thousand interacting cards. We present AutoTavern, a self-improving research system: specialised LLM research agents share a typed research memory, every claim passes verification gates, and live games close the loop, so the agents improve the simulator, opponents and policy and, from the same feedback, their own research process; a human sets goals and rules on ambiguous mechanics. In 20 days AutoTavern built TavernGym, the first faithful, full-content reproduction of the live game, and delivered 23 versions of TavernAgent, trained with PPO against opponents calibrated to real games. TavernAgent plays complete games on the live ladder by itself: its rating rose from 6,000 to about 15,000 and, on the public leaderboard's measure, its last 20 records average a placement of 1.60 with 80% first places, better than the server's ten highest-rated players on every statistic (2.34, 59%). TavernBench, 960 tasks in 40 scenarios, shows that the system the agents built is the strongest player: TavernAgent answers 91% of the play tasks against at most 63% for frontier LLMs, and decides in 0.03 s where they need up to 26 s; within the second a live turn allows, it keeps 90% and no LLM more than 26%. LLM agents build far better than they play. We will open-source TavernBench and TavernGym at https://github.com/MCG-NJU/AutoTavern.

Self-Evolution Reasoner: Continuous Optimization via Policy-Intrinsic Exploration and Exploitation

Wenhang Shi, Yiren Chen, Zhe Zhao, Jinhao Dong, Wei Lu, Xiaoyong Du — Tue, 27 Oct 2026 00:00:00 +0000

94% · New model 94%

Reinforcement learning with verifiable rewards (RLVR) enables large language models (LLMs) to continuously self-improve their reasoning capabilities. Compared to RLVR's scalar rewards, recent self-supervised on-policy distillation methods leverage denser supervision signals. However, both paradigms suffer from exploration bottlenecks, struggling to escape local optima when the policy fails to sample valid solutions. Relying on external demonstrations can bypass this issue; however, the off-policy nature hinders genuine self-improvement and the mismatched-distribution reasoning can not be fully exploited. In this work, we propose Self-Evolution Policy Optimization (SEPO), a framework that drives continuous reasoning capability evolution via guided exploration and in-distribution exploitation. Specifically, SEPO utilizes labels to elicit correct reasoning from the policy, unlocking the exploration potential. By treating current policy conditioned on this in-distribution reasoning as a self-teacher, SEPO distills the newly discovered solution back into the policy. Through this iterative cycle of policy-intrinsic exploration and exploitation, SEPO achieves robust and continuous capability evolution. Across scientific, tool-use, and complex logical reasoning domains, SEPO consistently achieves the best performance under identical training time or steps, demonstrating superior efficiency and convergence ceilings. Notably, SEPO maintains stable improvements even in scenarios where baselines severely underperform or fail entirely, such as on complex tasks or with small-scale models. Our code is available at https://github.com/Eric8932/SEPO.

Entropic Control for Homeostatic Optimization: Pre-empting Catastrophic Forgetting via Dynamic Feedback

Shangxi Wu, Xiaorong Yan, Xiaojun Meng, Yuxiang Zhang, Jiansheng Wei — Tue, 27 Oct 2026 00:00:00 +0000

94% · New model 94%, Efficiency 50%, Evaluation 43%

Supervised Fine-Tuning (SFT) of LLMs suffers from catastrophic forgetting, and prevailing mitigation strategies struggle to reconcile the plasticity-stability dilemma. To address this, we propose Entropic Control for Homeostatic Optimization (ECHO), a closed-loop control framework. By decomposing the KL divergence, we identify output entropy as a dominant observable component of KL variation. This enables us to leverage entropy as a computationally efficient online proxy to monitor forgetting risks in real-time, bypassing expensive reference-model inference. Consequently, ECHO employs an elastic controller that dynamically modulates the gradient weight of replay data via a proportional control law, using deviations from the pre-trained entropic equilibrium as an error signal. Empirical evaluations across 0.6B to 9B parameter architectures demonstrate that ECHO lies on, and in several comparisons extends, the observed plasticity-stability frontier and maintains high operational robustness against varying data mixing ratios.

MatOS: An Agent Operating System for Materials Research

Dingyang Lyu, Jing Gao, Zhonghao Chang, Jingjie Huang, Xiangran Cheng, Xin He — Tue, 27 Oct 2026 00:00:00 +0000

94% · Agents 94%, Evaluation 72%

Materials research relies on scattered and growing data, knowledge, and simulation tools, and a single study often spans multi-day workflows across literature analysis, computation, and experiment. Large language model agents can now retrieve materials data, operate simulation codes, and close design-synthesis-characterization loops with autonomous laboratories. However, extending these agents with new capabilities, keeping long-running tasks on track, and letting them learn from past tasks and user feedback remain difficult. Here we present MatOS, an agent operating system for materials research with three core components: (1) a layered scientific asset graph that organizes data, knowledge, tools, skills, roles, and workflows as reusable assets; (2) a stateful adaptive workflow that runs tasks as explicit execution graphs with scientific checks, so that they can be resumed and traced; and (3) evidence-driven asset evolution, which revises these assets from execution records and user feedback. On ResearchClawBench, MatOS improves on the previously best-performing agent by 11.15% with the same base model and extends the cost-score Pareto frontier. With a common base model, it outperforms Codex CLI and Claude Code, exceeding Codex CLI by 2.13% on MatTools and 4.17% on SciAgentGym. In eight research cases, MatOS narrowed more than 30 million candidate structures to 97 solid-electrolyte candidates, identified 89 halide variants with higher computed conductivity than their parent structures, linked the force errors of 11 pretrained interatomic potentials to local atomic environments, and supported two new methods that reduce errors by 3.1% and 13.6% relative to their baselines. These results show that organizing research capabilities as reusable, evolving assets enables agents to carry out long, multistage materials studies. Code is available at https://github.com/DingyangLyu/MatOS.

Birth, Death and Diff: Insertion-Based Sequence Generation with Corrections

Tue, 27 Oct 2026 00:00:00 +0000

94% · Efficiency 94%, New model 92%

Insertion-based masked diffusion models enable flexible-length, parallel generation by interleaving mask insertion with token unmasking. However, because their generation decisions are irreversible, aggressive parallel decoding can introduce errors that subsequent steps cannot correct, necessitating conservative decoding with more forward passes (NFEs). Moreover, they exhibit higher entropy than standard MDMs because they marginalize over multiple alignments during training and therefore require even more conservative decoding in the absence of a correction mechanism. To address this limitation, we introduce ${\delta}$MDM, which makes intermediate generations revisable by equipping an insertion-based model with substitution and deletion operations. During training, we expose the model to errors caused by aggressive unmasking and insertion, teaching it to correct its own mistakes. At inference time, we interleave insertion and unmasking with correction steps that can revise previously unmasked tokens and remove unnecessary masks. This mechanism makes aggressive decoding practical, improving both generation speed and quality. ${\delta}$MDM Pareto-dominates FlexMDM, a vanilla insertion-based MDM, on the speed--quality trade-off. On high-diversity unconditional text generation on OpenWebText, it halves the generative perplexity relative to FlexMDM at the highest inference budget. It also achieves lower generative perplexity at all inference budgets compared to MDM trained from scratch with variable-length padded sequences. On low-diversity conditional text generation on TinyGSM, it matches FlexMDM accuracy with half as many NFEs.

Improving the Chain-of-Thought Compressibility of Pruned Reasoning Models via Long--Short Contrastive Pruning

Tao Jing, Chen Kang, Feifei Zhai, Pengyuan Liu, Dong Yu — Tue, 27 Oct 2026 00:00:00 +0000

94% · Efficiency 94%, New model 84%, Evaluation 49%

Large reasoning models are costly to serve because they are large and because they generate long chains of thought (CoT). Weight pruning and CoT compression address these two costs separately, but they interfere in that pruning adds redundant reasoning on problems the model still solves and incomplete reasoning on those it no longer solves, and on harder problems makes the model less compressible by subsequent length-preference optimization. We trace this to a blind spot on each side: pruning criteria do not distinguish necessary from redundant reasoning, and preference-based CoT compression receives no training signal on problems a pruned model always or never solves. We propose Contrastive Set-Difference (CSD), which identifies neurons enriched in redundant reasoning by contrasting their activations on long and short correct solutions of the same problem and prunes them preferentially, and we extend preference pairs to the problems a pruned model always or never solves. Pruning 20\% of DeepSeek-R1-Distill-Qwen-7B, the combination uses 8--22\% fewer thinking tokens on four math benchmarks than pruning with reasoning-aware calibration (RAC) followed by the same compression, with accuracy within 0.5 points.

Efficient Proof Generation for Lean Statements with Multiple Candidate Answers

Pengzhen Lin, Ke Zhang, Maziar Raissi, Zhouxing Shi — Tue, 27 Oct 2026 00:00:00 +0000

94% · Efficiency 94%, New model 66%, Evaluation 58%

In formal systems such as Lean 4, efficiently finding answers to mathematical problems and generating verifiable proofs is an important task in automated mathematical reasoning. In the traditional way where an upstream solver provides multiple candidate answers, the common workflow requires searching for proofs for each candidate separately, incurring redundant computational cost. To this end, we propose a joint proof framework that integrates candidate answer selection, answer supplementation, and formal proof generation into a unified reasoning process. At its core is a trained prover that selects given or proposes new answers and generates a proof for at least one answer in a single call. Joint proof can also reuse generated proofs to check for faithfulness errors. For problems with a unique answer, if two distinct answers both obtain verifiable proofs under the same formal statement, this confirms the statement does not faithfully preserve the original problem semantics. On MATH-500, AMC 2023, and AIME 2024/2025/2026, our framework achieves the same formal proof success rate as per-candidate proof generation with about one sixth of the token cost, and identifies at least 6.1% of the formalized statements as unfaithful without requiring additional judges. Our model and code are available at https://github.com/Linppz/Gobble-Prover.

Error Diffusion Replaces Latent Weights: Training Binary-Weight Language Models on a Few Bits of Dithered State

Pierre-Olivier Bonin — Tue, 27 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 54%, Efficiency 31%

Binary-weight language models store one bit per weight, but training them keeps a full-precision latent copy of each weight plus Adam state, about 96 extra bits per weight, so a deployed binary model cannot be adapted on the device that runs it. We replace the latent weight with a 2-8-bit residual buffer quantized by stochastic rounding, which cuts the training state per weight from 97 bits to 3-9, up to 19x less. The resulting error-diffusion (ED) optimizer moves the residual-carrying loop of image halftoning onto the weight-update path. Each buffer accumulates the unrealized part of every update and flips the weight when the accumulated value crosses a threshold, and it works only when its quantization grid has a level exactly at that threshold. Under this threshold-alignment rule, 4-bit buffers match 8-bit buffers on a binary multilayer perceptron benchmark, and the same rule lifts a different latent-free optimizer, Bop, from a 10-point deficit at 4 bits to full-precision accuracy. On character-level transformers with 97% of parameters binary, ED trains within 5-11% of the latent-weight reference at a 2,000-step budget and within 1.7% at convergence. On scaling ladders over enwik8 and wikitext-103, up to 19M parameters on enwik8 (70M with the wikitext-103 embeddings), the latent-free loss stays within -2.3% to +2.9% of the reference at all six sizes once a per-step cap on the buffer update removes a launch instability; without the cap it trails by 10% at the largest tokenized size. A deployed binary model continues training under ED at one tenth of the latent training state, matching latent-weight continuation, and a fused CPU kernel completes the adaptation within a 275 MB training-memory budget, which latent-weight training cannot. A 250-step probe sets the flip learning rate at any new scale.

Prefetch-Steered Routing: Deterministic Cache-Aware Expert Selection for Offloaded Mixture-of-Experts Decoding

Siming Huang — Tue, 27 Oct 2026 00:00:00 +0000

94% · Efficiency 94%, Evaluation 70%

When a Mixture-of-Experts (MoE) model does not fit in GPU memory, batch-1 decoding is dominated by copying expert weights from host memory. Cache-aware routing reduces these copies by biasing the router toward resident experts, but combined with asynchronous prefetching it makes the model's outputs depend on transfer timing, so a harness that does not simulate that timing no longer evaluates the deployed model. We propose prefetch-steered routing (PSR): routing conditions on a deterministic logical cache, into which at most n experts predicted by the next layer's router are inserted per layer and token; physical copies affect only latency. On OLMoE-1B-7B with 50% and 25% of experts resident on an RTX 4090, PSR decodes 2.2-2.3x faster than exact routing with LRU and 25-30% faster than the Cache-Prior operating point, with mean benchmark differences of +0.1 and +0.5 points (95% lower bounds -0.8 and -0.3). In greedy generation its lead over Cache-Prior depends on the workload and reverses on HumanEval at 50% residency. On Moonlight-16B-A3B, PSR is 1.9x faster than exact routing with LRU at 37.5% residency and, at the latency of Cache-Prior, significantly better on three of eight benchmarks. On Qwen3.6-35B-A3B, where copies are cheap relative to computation, PSR keeps its advantage over Cache-Prior on the likelihood benchmarks but not its speed. Distillation-trained gates did not help. Speculative decoding with block-level expert selection was slower than PSR with a 1B autoregressive drafter on OLMoE, but with a block drafter on Qwen3.6 it was 2.1-2.9x faster than any autoregressive configuration with at most its GPU memory on the 64 prompts we timed.

Uncovering the Turing Test's Latent Signal: Indistinguishability as a Reward

Emanuele Rimoldi, Riccardo Neumarker, Tomaso Poggio, Daniel Mitropolsky — Tue, 27 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 73%

Can a model learn by trying to fool another model? We show that it can. Inspired by the Turing Test and its recent generalization to arbitrary distinguishers, we introduce a domain-based teacher-Turing-Test, in which a stronger teacher model serves as both the target of imitation and the distinguisher, and a weaker student model is trained to make the teacher attribute responses to itself. We use this self-attribution (or fooling) probability as a reinforcement learning reward, without ever providing the student ground-truth answers, nor requiring the teacher to generate responses. We test our approach independently in four domains with verifiable answers: CommonsenseQA (reasoning), GSM8K (mathematics), KodCode (programming), and Spider (SQL), training a Qwen2.5-1.5B student/imitator model using DeepSeek V4 Flash as teacher/distinguisher. Beyond improving at passing the Turing Test, the student improves on independent, held-out evaluations in all domains and runs. To probe why being mistaken for the teacher can improve capability, our analyses reveal that self-attribution is associated with correctness -- substantially more than formatting changes -- and that it rewards partial correctness. Our results demonstrate a new, simple teacher-student training paradigm with a private and/or high-cost teacher. They also suggest that indistinguishability need not merely measure behavior -- it can also be a mechanism for learning.

Collective Emergent Alignment: Measuring and Amplifying Interaction-Induced Reasoning Alignment in Multi-Agent LLMs

Boxuan Wang, Zhuoyun Li, Sarvapali D. Ramchurn, Xiaowei Huang, Yi Dong — Tue, 27 Oct 2026 00:00:00 +0000

94% · Evaluation 94%, Efficiency 56%, Hallucination 35%

Multi-agent large language model (LLM) systems can exhibit emergent behaviors through interaction, yet whether a similar phenomenon exists for reasoning alignment remains largely unexplored. We study whether agent interaction produces additional reference-grounded reasoning alignment beyond matched individual self-refinement, and introduce REMAP (Reference-Entailment-Matrix for Agentic Process), a framework that quantifies this effect through Semantic Coverage and interaction-specific alignment gains. Across diverse model families and two multi-hop reasoning benchmarks, we observe broadly positive interaction gains, particularly on reasoning content that depends on integrating evidence distributed across agents. These gains are consistently associated with independently measured peer-information recovery, while further analysis reveals composition-dependent heterogeneity effects and asymmetric information transfer across agent capability levels. Motivated by these findings, we further introduce a fact-level multi-agent communication framework that replaces full reasoning-trace exchange with compact, gap-aware, and selectively routed information. Our experiments show that compact fact exchange improves reasoning alignment and task performance while reducing communication overhead, whereas more selective communication exposes a trade-off between acquiring distributed information, retaining local content, and communication efficiency. Overall, our results show that collective emergent alignment can be quantified, linked to information-transfer patterns, and used to guide more effective multi-agent communication.

Adaptive Retrieval, Selective Answers: Evaluating Conformal Factuality in Agentic RAG

Muhammad Aimal Rehman, Chi-Kuang Yeh — Tue, 27 Oct 2026 00:00:00 +0000

94% · Hallucination 94%, Evaluation 89%, Agents 76%

Retrieval-augmented generation (RAG) can improve factual grounding without guaranteeing that every generated claim is supported. Previous work has studied claim-level conformal factuality in single-stage RAG and, more recently, multi-hop RAG. We examine its behavior in agentic RAG, where a language-model policy adaptively issues follow-up retrieval queries or terminates retrieval before response generation. We retain the retrieval-based relevance-scoring formulation and post-generation conformal filtering and evaluate six model-dataset configurations: GPT-6 Luna and Llama 3.1 8B on HotpotQA, Natural Questions, and TriviaQA. Each configuration uses 60 fixed questions, 100 seeded 30/30 calibration-test partitions, five conformal target levels, and a no-conformal-prediction baseline. Across the six configurations, higher target levels are accompanied by higher observed full-support rates and reduced answer availability. At the 95% target, non-empty response coverage ranges from 17.23% to 60.17%, claim retention from 10.72% to 43.16%, and full support among non-empty outputs from 89.14% to 95.56%. We formulate the adaptive retrieval trajectory explicitly and characterize the reliability-informativeness trade-off when conformal factuality is applied after agent termination.

Correlation-Aware Exploration for Long-Trace Reasoning

Junxing Hu, Zicheng Zhang, Zhiqian Zhang, Ai Han, Yifeng Zhang, Pengzhang Liu — Tue, 27 Oct 2026 00:00:00 +0000

94% · New model 94%, Efficiency 58%

Reasoning models trained with Group Relative Policy Optimization (GRPO) now produce chains of thought thousands of tokens long, and recent studies reveal that such training is prone to policy collapse. Entropy regularization is the standard remedy, but it works by pushing probability mass away from the few tokens that carry it and toward the vast low-probability tail of the vocabulary, and it applies this force once for every token of the reasoning trace. This puts the bonus in direct conflict with the entropy decay that convergence requires, and makes its coefficient difficult to set in the long-trace regime. To resolve this tension, we propose Info-GRPO, an information-theoretic framework that cultivates correlation between the policy and a latent prior. Rather than raising entropy everywhere, Info-GRPO builds on the mutual information between a latent variable and the trajectory it induces: augmenting prompts with latent seeds during training lets the model explore diverse policies correlated with the prior, while the seed-conditioned entropy is guided toward convergence. Diversity is therefore carried across whole reasoning traces rather than by per-token randomness, and exploration no longer fights entropy reduction. Experiments show that Info-GRPO outperforms vanilla GRPO and entropy-regularized GRPO across diverse reasoning benchmarks. On AIME24, it improves $Avg@8$ over GRPO by 3.75%, 1.66%, 4.16%, and 4.17% on four backbones, and it reaches the same majority-vote accuracy with $8\times$ fewer samples. Code and models will be publicly released.

Improving Reasoning with Rollout Agreement for Group-Based Policy Optimization

Sohyun Son, Yerim Hong, Jaeyun Park, Mujeen Sung — Tue, 27 Oct 2026 00:00:00 +0000

94% · New model 94%, Hallucination 71%, Evaluation 61%

Recent methods for Reinforcement Learning with Verifiable Rewards (RLVR) use rollout diversity to improve reasoning performance or training efficiency. Despite these gains, the extent to which differences among correct-answer rollouts reflect useful alternative reasoning or unresolved errors remains poorly understood. We investigate this question through a blind paired analysis of a model trained with Group Relative Policy Optimization (GRPO), comparing the highest- and lowest-agreement correct-answer traces for the same problem. Across 126 pairs, low-agreement traces have more spurious reasoning than high-agreement traces (35% vs. 22%), including internal inconsistencies and misuse of problem conditions. Based on these findings, we propose RoAR (Rollout-Agreement Reward), which adds reasoning-level agreement to GRPO's rewards for verified-correct rollouts. RoAR computes this signal from existing on-policy rollout groups, requiring no human-written rationales, additional sampling, or separately trained process evaluator. On five mathematical reasoning benchmarks, the 1.5B GRPO-RoAR model reaches 55.5% mean pass@1 under zero-shot evaluation using 7K training examples, exceeding GRPO and GRPO-CARE by 6.0 and 5.5 percentage points. A separate blind audit of 100 problems answered correctly by all four 1.5B models finds spurious rates of 18% for GRPO-RoAR, 32% for GRPO, 27% for GRPO-CARE, and 37% for DRA-GRPO. Our findings identify reasoning validity as an important consideration when using rollout differences to guide learning. Our code is available at https://github.com/ldi-kyunghee/RoAR.

Reinforcement Learning Done Right for Explicit Confidence Targets

Zefeng He, Yichen He, Yuan Lin, Yang Yang, Yu Cheng — Tue, 27 Oct 2026 00:00:00 +0000

94% · New model 94%, Hallucination 73%, Evaluation 34%

Large language models (LLMs) can produce plausible but incorrect answers when they are uncertain, which can mislead users and undermine model reliability. Training models to abstain with reinforcement learning is a natural approach, but most existing methods use a fixed error penalty and therefore cannot adapt their behavior to different confidence requirements at inference time. We study \emph{explicit confidence targets}, where a user specifies a target $p$ in the prompt and the model directly chooses whether to answer or abstain, with the goal of achieving an accuracy of at least $p$ among answered questions. However, learning to respond to different confidence targets with a single model has proven difficult, and prior attempts have failed to distinguish different targets. We show that this failure comes from two problems: the reward signal that distinguishes different values of $p$ is often lost during optimization, while the original error penalty grows without bound as $p$ approaches one. We propose RLEC (\textbf{R}einforcement \textbf{L}earning with \textbf{E}xplicit \textbf{C}onfidence Targets), which preserves the learning signal needed to distinguish confidence targets, introduces a stable reward family that keeps the desired answer condition unchanged, and uses reward scheduling to balance effective learning and training stability. Together, these designs enable the first successful training of a single LLM for explicit confidence target control. Extensive experiments across diverse benchmarks and model scales show that RLEC largely preserves general capabilities and achieves competitive calibration performance. RLEC also supports fine-grained target control without an external decision system or threshold calibration after training. By making explicit confidence targets part of the trained policy, RLEC opens a new and more direct path toward behavioral calibration.

DTCA: Denoising Token Credit Assignment for Reinforcement Learning in Diffusion Language Models

Hongze Chen, Miao Pan, Lizhou Liang, Qinfeng Li, Xinyin Ma, Yongliang Shen — Tue, 27 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 31%

Diffusion language models (DLMs) generate text through iterative denoising, enabling bidirectional context modeling and parallel token refinement. Recent studies have applied reinforcement learning (RL) to DLMs and improved their reasoning performance. However, the iterative denoising process makes credit assignment more challenging, as final outcome rewards provide limited information about how different tokens contribute to the final output. In this paper, we propose Denoising Token Credit Assignment (DTCA), a framework for token-level credit assignment in reinforcement learning for diffusion language models. DTCA estimates token contributions by measuring the improvement in prediction confidence of remaining masked positions after token commitment and groups these signals according to part-of-speech (POS) categories. These contribution weights are used to select informative denoising timesteps and assign token-level learning signals during RL training. DTCA further calibrates these learning signals according to token-level entropy changes induced by policy updates. Experiments on LLaDA-8B-Instruct show that DTCA consistently improves mathematical reasoning performance across different generation budgets, with gains of up to 9.0 points on MATH500 and 5.3 points on GSM8K over the base model.

Shared Context, Separate Engines: Streaming Vision--Language Inference on Edge SoCs

Yilong Li, Chengpo Yan, Yijing Zeng, Hanlei Wu, Mingjun Song, Suman Banerjee — Tue, 27 Oct 2026 00:00:00 +0000

94% · Efficiency 94%

Edge SoCs pack CPUs, GPUs, and NPUs, yet vision--language model (VLM) runtimes keep generation on one engine and leave the others idle. Streaming VLMs expose further parallelism: new frames arrive while earlier queries are still being answered. Engines differ in compute and memory access and contend when they co-run, so no single placement of stages wins across SoCs and workloads. Backend-local KV state keeps other engines from continuing the work. \name coordinates heterogeneous accelerators to execute stateful VLM inference concurrently. It separates a stream's versioned context from each engine's KV state, so any engine can continue the stream by reusing, importing, or reconstructing that state. Stateful Streaming Generation ingests new frames on one engine while another answers, and Concurrent Sequence Generation serves different ready requests in parallel. State-aware dispatch sends each request to the engine expected to finish it first. Each answer reads a fixed, published version of its history and matches its serial reference. \name beats the strongest single-engine baselines on Apple M4 Max, Rockchip RK3588, and Qualcomm QCS6490, even when they reuse context. On two RK3588 video streams, it delivers $19.6\%$ more throughput than the best single engine with context reuse and $27.2\%$ more than parallel engines without shared state. Cross-engine KV import cuts first-token latency by a factor of $4.6$ over prefix recomputation. Stateful Streaming Generation delivers up to $2.48\times$ the throughput of serial execution, and state-aware dispatch cuts query completion time by up to $72.7\%$.

GPDPO: Learning from Reward Conflicts via Group Pareto-Dominance Policy Optimization

Zixuan Xu, Haoxuan Chen, Jiangxin Sun, Jian-Fang Hu — Tue, 27 Oct 2026 00:00:00 +0000

94% · New model 94%

Multi-reward reinforcement learning offers a promising framework for aligning large language models (LLMs) with diverse objectives, but conflicting reward signals can induce ambiguous update directions and hinder stable optimization. Existing approaches address such conflicts through reward reweighting, gradient coordination, or rollout filtering, yet they primarily operate on per-response or group-level optimization signals and underexploit relational information among sampled responses. As a result, conflict-filtered rollouts may still contain useful supervision that is discarded during training. We propose Group Pareto-Dominance Policy Optimization (GPDPO), a multi-reward policy optimization method that recovers pairwise supervision from conflict-filtered rollouts. GPDPO identifies Pareto-dominance relations among sampled responses and converts them into auxiliary preference signals, which are jointly optimized with the original policy objective. This enables more effective reuse of existing rollouts without additional sampling or reward evaluation. Experiments on API-Bank across multiple model scales and reward configurations show that GPDPO consistently outperforms competitive multi-reward optimization baselines. Further analysis demonstrates that GPDPO reduces both group-relative and pairwise reward conflicts, supporting the effectiveness of exploiting relational supervision for multi-reward alignment. Code will be released.

Multi-Agent Retrieval-Augmented Causality Assessment (MARCA) of Individual Case Safety Reports in Pharmacovigilance

Daisy Rae Lopez, Angela Boccia, Maurizio Sessa — Tue, 27 Oct 2026 00:00:00 +0000

93% · Agents 93%, Evaluation 75%

Introduction: Increasing volumes of individual case safety reports (ICSRs) challenge expert-led causality assessment in pharmacovigilance. We developed MARCA (Multi-Agent Retrieval-Augmented Causality Assessment), a four-agent framework for Naranjo-based assessment, and evaluated agreement with human experts and exploratory differences between non-RAG and RAG-labelled scores. Methods: MARCA combined scientific literature retrieval, drug-label assessment, objective-evidence classification, and clinical causality reasoning using Gemini 2.5 Pro and DeepSeek-V3.2. Agent-generated rationales informed human-assisted scoring. A consensus expert reference standard covered 215 ICSRs and 1,562 drug-event combinations (DECs) from FAERS (Q1-Q2 2025). Agreement was measured using percentage agreement and Cohen's kappa. Results: Question-level agreement was 96.8% for Naranjo Q2 (kappa = 0.491; 95% CI 0.265-0.663), 85.4% for Q5 (kappa = 0.269; 95% CI 0.180-0.340), and 96.8% for Q10 (kappa = 0.922; 95% CI 0.885-0.951). Final causality classification agreement was 79.4% (95% CI 71.8-85.3; kappa = 0.587, 95% CI 0.450-0.690). In a retrospective subset of 60 DECs, agreement was higher for RAG-labelled than non-RAG scores for Q5 (83.3% vs 65.0%) and Q10 (100.0% vs 18.3%). Discussion: MARCA-assisted assessment showed substantial variation in agreement across Naranjo questions, with lower concordance for alternative causes. Retrospective RAG comparisons are exploratory and do not establish prospective effectiveness.

StateWise: Teaching Language Models to Track and Obey Evolving Constraints in Multi-Turn Dialogue

Samarth Varshney, Ritu Yadav — Tue, 27 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 79%

Users of conversational assistants rarely state all requirements up front: they add constraints, amend them, withdraw them, digress, and return. Instruction-tuned LLMs fail here in two ways: active constraints stop being honored as turns accumulate (persistence failure), and withdrawn constraints keep being enacted (relapse). The only existing repair we know of is an external ledger that re-injects a compiled specification every turn. We propose StateWise, a self-improving approach that teaches a single LLM to maintain its own revisable constraint state, a typed state with tombstones for revoked constraints, so persistence and relapse are measured by executable checkers and ground-truth states are compiled for free. StateWise fine-tunes the LLM on two complementary skills, summarizing a dialogue into its constraint state and responding conditioned on that state, then self-trains on its own trajectories, keeping only turns that pass an exact round-trip check against the compiled state. This teaches 7-8B models to track the state almost perfectly but, surprisingly, not to obey it: on Llama-3.1-8B every supervised variant roughly doubles relapse. We therefore add a preference stage on self-generated relapse contrasts, where the rejected response enacts a revoked rule and the chosen one passes the round-trip check. On two backbones with held-out constraint families and unseen phrasings, StateWise cuts relapse by 8-22 points relative to supervised self-training, lifts conversation success by 8-19 points over the backbone, and beats an oracle compiled-ledger baseline on relapse with no inference-time scaffolding. We also characterize the recipe's cost on general instruction-following benchmarks and study weight interpolation as a mitigation.

EnvLunch: The Environment as a Free Lunch for Learning to Interact

Wei Liu, Min Gao, Yali Du, Yulan He — Tue, 27 Oct 2026 00:00:00 +0000

93% · New model 93%, Agents 58%, Evaluation 41%

For agentic data analysis, language-model agents often require a large number of annotated tasks to learn how to interact effectively with executable environments (i.e., databases). While high-quality annotated tasks require heavy manual effort, an executable environment already provides rich interaction structure and verifiable outcomes. In this sense, the environment itself offers a free lunch, an additional set of verifiable tasks that require no human annotation. We propose EnvLunch, a warm-up stage that turns this environmental signal into agent training without human-written or model-generated task descriptions, before downstream agentic reinforcement learning. EnvLunch samples valid action trajectories, executes them to obtain final observations, and trains an agent to recover a functionally equivalent trajectory given only its final observation. This label-free warm-up enables the model to outperform the baseline by over 9 points on average across all downstream training budgets, and with few labels, the baseline needs 8 to 32 times as many annotated examples to match the accuracy that the EnvLunch model reaches at its smallest training budget. Letting the environment alone define the task yields a larger gain than having the model write its own tasks over the same environments, and this warm-up also improves fine-tuning on benchmarks beyond data analysis. EnvLunch therefore saves annotated tasks and provides a better starting point for fine-tuning agentic models when such tasks are limited.

When Personas Become Harmful: Guiding LLMs with Global Citizenship Values Alignment

Siqi Shen, Lajanugen Logeswaran, Moontae Lee, Honglak Lee, Rada Mihalcea — Tue, 27 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, New model 83%

LLMs are increasingly being used by users from diverse demographics with pluralistic values and distinct backgrounds. Maintaining safety while respecting value-laden personas is uniquely challenging because persona-conditioned requests are often not overtly malicious, unlike those in existing safety benchmarks. Alignment methods aimed at adversarial requests therefore may miss the harm that emerges from persona use, leaving models to either refuse legitimate value expression or blindly follow unsafe personas. We construct a dataset of 11k controversial topics paired with value-based personas mapping to value dimensions from the Schwartz Value Survey. Through an agentic pipeline that iteratively refines the topic and persona, we surface and collect harmful responses from well-aligned commercial LLMs with a more than 47% success rate. We use UNESCO's global citizenship principles as a minimal common ground for distinguishing valid value expressions from harmful behaviors, and introduce a controllable alignment model that lets users balance persona adherence and safety at inference time. The resulting model is safer and more persona-adherent than baselines on most samples.

INTRA: Interleaved Non-contiguous Token spaRse Attention

Yuxuan Xia, Wenbo Lu, Shaoyi Zheng, Shengjie Wang — Tue, 27 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 36%

Transformers achieve strong performance across modalities but are bottlenecked by the quadratic cost of attention. Sparse attention can reduce this cost, but effective sparsity must preserve useful long-range communication and map well to GPU hardware. We introduce INTRA--Interleaved Non-contiguous Token spaRse Attention--a hardware-aware static sparse attention design based on a relay-communication view under Computational Query Set (CQS) constraints. Rather than directly covering all token-to-token interactions within each sparse layer, INTRA uses complementary sparse groups to provide short mediated communication paths across layers. Scatter forms relay groups that sample broadly across local neighborhoods, while Gather reconnects these groups in subsequent layers. Under a partition model, this structure guides CQS-compatible sparse patterns with global reachability. Our CQS-aware FlashAttention-style kernel computes structured indices at runtime, loads selected non-contiguous tokens, and packs them densely for computation. On an A100, INTRA achieves a $1.53\times$ speedup in $2K$ FLUX.1-dev generation, with better measured image quality than evaluated CLEAR configurations after LoRA self-distillation. On LLaMA-3.1-8B, it improves long-context prefill efficiency with strong teacher-forced retrieval performance under matched LoRA adaptation. Attention-only benchmarks further show up to $8.4\times$ and $10.4\times$ speedups over MoBA and density-constrained NSA, respectively.

SRA: Selective Residual Adaptation for LLM Reasoning Enhancement

Zhenxing Dou, Kun He, Mingyang Yi, Yong Liu — Tue, 27 Oct 2026 00:00:00 +0000

93% · New model 93%, Efficiency 34%, Evaluation 30%

Knowledge distillation transfers reasoning capabilities from large language models to lightweight students. However, distilled small models still struggle with multi-step reasoning. We observe that the distribution divergences between teacher and student models along reasoning paths are highly sparse, and target tokens almost always fall within the student top-$K$ candidate set. Based on this finding, unlike approaches updating student backbones or requiring runtime teacher computation, we propose Selective Residual Adaptation (SRA), a plug-and-play framework that internalizes localized corrections. Concretely, SRA employs a Discrepancy Gate to identify divergence positions alongside a Candidate Residual Adapter for top-$K$ logit calibration to correct the mismatch. Theoretically, we establish error bounds ensuring strict sequence level performance gains. Across mathematical benchmarks on Qwen3 models from 0.6B to 4B, SRA consistently enhances distilled student models, raising average Pass@16 by up to 4.37 percentage points while preserving general capabilities.

DoubleFree:Exploiting Exploration History for Rollout-Free Multi-Task LLMs Training

Yingping Wang, Min Cai, Chenyu Yuan, Xi Ye, Zhijiang Guo, Jiayi Huang — Tue, 27 Oct 2026 00:00:00 +0000

93% · New model 93%, Efficiency 89%

Multi-task post-training can develop domain specialists through independent reinforcement learning (RL) and consolidate their capabilities through multi-teacher on-policy distillation (MOPD). However, standard MOPD generates on-policy student responses and keeps multiple teachers online throughout consolidation, introducing substantial system costs. We view this pipeline as domain-wise exploration followed by cross-domain exploitation. Domain RL produces not only the final specialists, but also the trajectories collected throughout their training. Existing methods primarily use the former while leaving the latter underutilized. We introduce DoubleFree, which reuses these historical trajectories for rollout-free consolidation and precomputes teacher supervision for online-teacher-free student training. To understand why this works, we analyze when updates from historical trajectories can approximate on-policy distillation updates and how combining trajectories across exploration stages can better support an evolving student. DoubleFree requires only minimal changes to existing post-training frameworks. Across math, code, and instruction following on Qwen3-4B, Qwen3-8B, and SmolLM3-3B, DoubleFree achieves $2.99$-$5.75\times$ consolidation speedups and reduces GPU-hours by $80.9$-$90.1\%$, while maintaining comparable multi-domain performance.

ScaleRLT: Exact Log-Probabilities and Fast Deferred Rollouts for Reinforcement Learning on Looped Transformers

Dezhen Lu, Zixiang Ni, Runguo Li — Tue, 27 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, New model 76%, Evaluation 51%

Looped language models such as Ouro reuse one block of layers several times per token. A rollout engine can batch the deeper loops of several response positions and run them only when a deep readout is due. This deferred schedule is fast, but the learner of an on-policy RL algorithm recomputes the sampled tokens' log-probabilities with a dense pass, and the two computations differ in floating point: in FP32 at 4096 tokens about 18% of the sampled tokens miss a $10^{-5}$ same-weight gate, and with bf16 weights nearly every token does. We use a hybrid learner log-probability, $\ell=\ell_{\mathrm{dense}}+\mathrm{sg}(\ell_{\mathrm{sched}}-\ell_{\mathrm{dense}})$, where $\mathrm{sg}$ denotes stop-gradient. Its value is a no-grad replay through the rollout's own schedule, and its gradient is the dense learner's. The replay reproduces the rollout bit for bit in all 167 FP32 and 28 bf16 real-weight cells we ran on three platforms, including rollouts whose batch shrinks at stop tokens, and the gradient has the same FP32 error as other dense implementations in all 102 pre-registered comparisons from 96 to 4096 tokens; the full GRPO loss of a training step matches a float64 reference to $3.1\times10^{-4}$. On one GPU, deferral raises rollout throughput by $1.47$ to $1.56\times$, and by $1.18$ to $1.54\times$ once every step runs from CUDA graphs, bit for bit. In a two-node GRPO loop, one-wave KV admission and CUDA graphs shorten an update by $2.71\times$ with the gate bitwise at every update, and four nodes keep the update time within 2% at the same per-GPU load. In 100-update runs held-out GSM8K accuracy rose from 23% to 82% with the mixed-depth readout that deferral relies on and from 34% to 90% with the native readout, which reached a higher accuracy in less time. For looped models beyond one node we analyse a ring mapping of layers to GPUs and measure its communication paths.

Heard Enough to Act: Early Tool Calling in Spoken Dialogue Systems

Yizhou Peng, Chen Chen, Donghang Wu, Hexin Liu, Eng Siong Chng — Tue, 27 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, Agents 91%, New model 83%

Latency-sensitive spoken agents must know not only which tool to call, but also when they have heard enough to act. Waiting for the turn to complete delays execution even when the request is already actionable. However, neither existing spoken agents nor spoken dialogue corpora define the earliest point at which a request becomes actionable. We introduce a synthetic spoken-dialogue pipeline that provides earliest-actionable-point supervision, marking when the observed speech grounds a tool and its required arguments. Through controlled task generation, speech synthesis and alignment, and gated annotation, we construct 26,694 dialogues totaling 786 hours across 48 domains, with disjoint training and test argument values. We additionally create an audio rendering of $\tau^2$-bench for external evaluation. We train an omni-modal language model on this corpus, enabling causal, chunk-level decisions and partial-transcript generation for downstream call completion. The model achieves tool-state macro-F1 scores of 0.48 and 0.47 on held-out dialogues and the external benchmark, respectively. For correct calls, the estimated mean lead over a turn-end cascade, using the same formatter, is 1.8-2.0 seconds in-domain and 1.1 seconds on $\tau^2$-bench. These results demonstrate a data-driven path toward spoken agents that act while listening. We will open-source the pipeline and the training scripts.

Multiscale Policy Optimization for Long-Horizon LLM Agents

Tao Ren, Hui Yang, Jia Deng, Tingyang Chen, Qiyao Wang, Shuo Tang — Tue, 27 Oct 2026 00:00:00 +0000

93% · New model 93%, Agents 86%, Evaluation 30%

Long-horizon LLM agents interact with their environments through complete assistant turns while generating each action token by token. We observe that critic values vary primarily across turns, motivating optimization anchored at turns with finer credit allocation among tokens. We propose Multiscale Policy Optimization (MPO), combining Multiscale Advantage Estimation (MAE) and turn-level importance sampling to align credit assignment and policy updates at the turn scale. Extensive experiments across search, tool-use, and automation tasks with 4B and 30B-A3B models demonstrate consistent gains over strong RL baselines.

SimPOTrans: Self-Adaptation for Offline-to-Simultaneous Speech Translation via Reinforcement Learning

Haotian Tan, Hiroki Ouchi, Sakriani Sakti — Tue, 27 Oct 2026 00:00:00 +0000

93% · New model 93%, Efficiency 52%

Simultaneous speech translation (SimulST) aims to translate ongoing speech with low latency while preserving translation quality. Recent studies have adopted large language models (LLMs) for SimulST and formulated it as a multi-turn dialogue task for efficient generation, achieving state-of-the-art performance. However, existing dialogue-style SimulST approaches typically rely on specialized training schemes to acquire simultaneous translation behavior, either using synthesized read/write trajectories, which may provide suboptimal training signals, or jointly optimizing translation quality and latency, whose competing objectives may cause optimization instability. In this paper, we introduce SimPOTrans, a reinforcement learning (RL) framework that enables offline speech translation models to self-adapt to dialogue-style SimulST. Through a simple two-stage RL post-training pipeline, SimPOTrans rewards an offline model for exhibiting translation behaviors beneficial to simultaneous scenarios when given partial speech inputs, thereby enabling offline-to-simultaneous self-adaptation without synthesized read/write trajectories or explicit latency optimization. At inference time, SimPOTrans requires no external decision policy and enables flexible quality--latency control by simply adjusting the chunk size of speech input. Experiments on the ACL 60/60 and MCIF benchmarks across three language directions (En$\rightarrow$De, En$\rightarrow$Zh, and En$\rightarrow$Ja) demonstrate that SimPOTrans outperforms several strong baselines, achieving a superior quality--latency trade-off.

Did the Model Use That? Counterfactual Training Boosts Self-Attribution in Language Models

Zikai Zhou, Jiacheng Sang, Yilin Xu, Yufei Jin, Monica S. Lam — Tue, 27 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 73%

A fundamental challenge for explainable and accountable language models is revealing how external information influences their answers. This paper examines whether models are self-attributive: that is, whether they indicate when -- and only when -- their responses to questions are influenced by information provided alongside the question. Furthermore, we explore how self-attribution can be improved through counterfactual training, which explicitly teaches models to distinguish between answers that do and do not depend on supplied information. We found that most of the eleven frontier models rarely attribute influential information in their answers, and the rest attribute non-influential information. Training acknowledgment into reasoning does not fix this: the acknowledgment does not reach the final answer. We introduce CREDIT, which directly supervises final-answer self-attribution with counterfactual labels, without the need for human annotation. CREDIT substantially improves attribution in four open models, including two distilled reasoning models, and outperforms reasoning-level fine-tuning and prompting baselines. Among answers that agree with a user's suggestion, CREDIT distinguishes those changed by the suggestion from those the model would have given anyway. Gains transfer partly to unseen information types and prompts containing sources that support conflicting answers.

Belief-Conditioned Multi-Agent Actor Critic for Decentralized LLM Collaboration

Tianle Chen, Shuo Liu, Christopher Amato — Tue, 27 Oct 2026 00:00:00 +0000

93% · Agents 93%, New model 78%

Recent work has explored optimizing LLM collaboration through Multi-Agent Reinforcement Learning (MARL), but many approaches still rely on predefined workflows, limiting efficiency and scalability. Decentralized LLM collaboration is challenging because agents execute independently with limited communication, share a joint objective, and must infer useful information from communication traces. In this paper, we propose BCMAAC, a Belief-Conditioned Multi-Agent Actor-Critic method that maintains beliefs over contextual information and uses them to facilitate training. Experiments on Minecraft game-playing, web searching and tool-using tasks show that BCMAAC improves task performance and collaborative behavior compared to naive actor-critic baselines. The improvements are particularly pronounced in multi-turn tasks with richer contextual information, where belief modeling provides a clear advantage.

Faithful Explanations under Missing Modalities in Multimodal Dementia Prognosis

Firuz Kamalov — Tue, 27 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, Hallucination 93%

Large language models are increasingly used to turn the feature attributions of multimodal prognostic models into narratives for clinicians, yet costly modalities such as amyloid PET and cerebrospinal fluid biomarkers are often missing, and whether these explanations stay faithful under missing evidence has not been tested. We separate two failure points: phantom attribution, in which the model credits modalities that were never observed, and unfaithful narration, in which the report omits, cites or misjudges missing modalities. Using 307 participants of the Alzheimer's Disease Neuroimaging Initiative with complete data, we masked every subset of four costly modalities, predicted 36-month conversion from mild cognitive impairment to Alzheimer's disease, and restored the masked data to obtain interventional ground truth. All five missing-data strategies assigned 12% to 23% of attribution to unobserved modalities, and up to 45% with four absent; missingness-aware strategies minimised it at no cost in accuracy. Across 51,200 reports from four language models and four prompting regimes, naming absent modalities in the prompt raised acknowledgement from 48% to 88% and cut phantom evidence twenty-fold, but advice on whether acquiring a missing test would change the prediction was correct in only 27% to 41% of cases until checked against executed restorations, which raised accuracy to 69% for open-weight narrators and 99.9% for a proprietary model. Higher phantom attribution was associated with more phantom evidence and fewer faithful direction claims, and direction fidelity, measured by a calibrated language-model judge, depended on narrator scale rather than prompting. Narrators told to name absent modalities sometimes declared observed ones absent, an error that a structured template eliminated. We release the masking protocol and evaluation code.

S3AC: Stabilized, Selective, and Scalable Actor-Critic for LLM Reasoning

Shuo Shen, Guangran Cheng, Chengqi Lyu, Xiangsen Wang, Yanan Sun, Wenwei Zhang — Tue, 27 Oct 2026 00:00:00 +0000

93% · New model 93%, Efficiency 41%, Evaluation 31%

Critic-free methods are widely used in reinforcement learning with verifiable rewards (RLVR) for long chain-of-thought (CoT) reasoning, avoiding difficult and costly critic training. Common alternatives lack token-level credit assignment and typically rely on grouped rollouts to estimate prompt-specific baselines. A critic can address both limitations, but long-CoT critic training is hindered by insufficient coverage of trajectories with diverse lengths and early gradients dominated by the newly initialized value head. We introduce S3AC (Stabilized, Selective, and Scalable Actor-Critic) for LLM reasoning, built around a practical critic-training recipe. The trained critic provides token-specific advantages and prompt-specific baselines. Using the prompt-level baseline to estimate difficulty, we propose Difficulty-Aware Prompt Sampling (DAPS). DAPS combines this estimate with outcome history to select prompts before generation, reducing rollout waste without offline difficulty initialization. S3AC improves group-rollout performance across dense and mixture-of-experts models from 1.5B to 397B parameters, outperforming a modern GRPO variant (GRPO-X). At 35B, it sustains single-rollout training while tested critic-free baselines deteriorate after initial gains.

MiRL: Error-Targeted Reinforcement Learning for Correcting Terminal Agent Misbehaviors

Runchu Tian, Zaid Khan, Zun Wang, Justin Chih-Yao Chen, Mohit Bansal — Tue, 27 Oct 2026 00:00:00 +0000

93% · New model 93%, Agents 67%, Efficiency 54%

Error-targeted learning gives a legible way to improve a model: identify specific weaknesses of the model, train the model to fix these weaknesses, and check whether the weaknesses are fixed. We study how to apply error-targeted learning to terminal agents, which solve long-horizon tasks specified in natural language by interacting with a command shell. Existing error-targeted methods are designed for single-turn tasks, and existing analyses of agent failures are used to evaluate and debug agents rather than to train them. We propose Misbehavior-driven RL (MiRL), which first profiles the errors of a policy to find generalizable misbehaviors: kinds of wrong steps that recur across many environments. MiRL then trains the policy with RL, starting rollouts at the steps where these misbehaviors occur and rewarding a rollout only if it passes the tests and avoids the misbehavior. On Qwen3.5-4B, eight misbehaviors account for 52.1% of wrong steps. After training with MiRL, the policy repeats the measured misbehaviors 43% to 57% less often on average on unseen prefixes. MiRL outperforms TMax-4B on Terminal-Bench Lite (51.8% vs. 47.4%) and matches it on Terminal-Bench 2.1 (21.3% vs. 20.6%), while using 35% of its rollouts. These results show that error-targeted learning can be applied to realistic, long-horizon terminal tasks.

FOCUS: Active Critic Learning for Long-Horizon Language Model Agents

Tue, 27 Oct 2026 00:00:00 +0000

93% · New model 93%, Agents 78%, Evaluation 42%

Sparse outcome rewards make it difficult for long-horizon language-model agents to assign credit to earlier actions. Learned critics provide value estimates for intermediate states, but these estimates can be inaccurate, while estimating intermediate values with repetitive continuation samples from points within a trajectory incurs substantial costs. To mitigate these issues, we introduce FOCUS, an RL method that uses selective branching to help the critic network learn its value estimator. The rollout controller facilitates active learning for the critic by sampling continuations from states that are expected to most reduce the critic's estimation error. These refined value estimates then score the segment between consecutive states with a TD($\lambda$) advantage, while the sampled evidence supplies training targets for the critic. FOCUS supports long-horizon agentic reinforcement learning through a modular context-management interface that decouples context management from rollout allocation. Our experiments show that using compaction-induced segment boundaries for value estimation and credit assignment, FOCUS outperforms GRPO and VinePPO with matched per-step generation budgets on legal reasoning, deep search, and code localization benchmarks, improving held-out accuracy at Qwen3.5-9B scale by 7.2 points over the strongest baseline on BrowseComp-Plus (55.4 vs. 48.2) and by 4.1 points on DeonticBench (56.5 vs. 52.4).

Two Uncertainties Are Not One: Disagreement versus Decision-Risk in Cheap Structured Content Moderation

Karan Yadav — Tue, 27 Oct 2026 00:00:00 +0000

92% · Evaluation 92%, Efficiency 52%

Content moderation compresses a graded, contested human judgment into a binary act (leave up / take down) over billions of items, at a scale where running a frontier LLM on every post is uneconomical. A new class of frozen "System One" models returns typed, calibrated decisions instead of text at ~USD 0.000025 per call and ~100 ms. We run the first independent audit of one such model on 142,658 human-labeled comments, and benchmark it against 14 frontier reasoning models (the GPT-5.x, Claude, Gemini 3, and Grok families) on a shared probe set. Two findings stand out. First, the cheap model is accuracy-competitive: its zero-shot classification AUROC (0.83) is not statistically separable (paired bootstrap, n=56; a failure to reject parity, not proof of equivalence) from 13 of the 14 frontier systems while costing 1-3 orders of magnitude less, and it is no worse calibrated than several of them. Second, "confidence" conflates two uncertainties that come apart: human disagreement (where annotators split on satire and coded language) and decision risk (where the automatic action is costly-wrong). Under an asymmetric moderation loss these call for opposite escalation signals, so no single confidence number serves both. Because the model prices extra questions at nearly zero, decomposing the judgment into parallel sub-questions and reading their spread recovers a human-disagreement signal that the model's native confidence misses -- but this remedy is specific: it helps the weakest, disagreement-blind models and only when the decomposition aligns with the sources of disagreement; a specialist toxicity model and the strongest LLMs do not need it, and the effect is distribution-sensitive. We will release the instrument, the cached runs, and the analysis.

When the Weakest Language Keeps Moving: ARMOR-DP for Multilingual BabyLM

Hanyu Liao, Yijing Chen, Tao Ji — Tue, 27 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 65%

Multilingual BabyLM systems usually fix language sampling proportions even though the weakest language can change during training. We present ARMOR-DP, a modular framework that diagnoses imbalance arising from representation, exposure, capacity, and objective pressure, and uses per-language validation loss to update sampling. Controlled experiments show that equal character budgets create a severe representation artifact for Chinese; byte-level encoding removes much of it, while static Chinese protection can subsequently make English the weakest language. The dynamic controller tracks these transitions and achieves a better score--balance tradeoff than fixed protection. On official evaluation, curriculum-dynamic checkpoints outperform a matched static Chinese-protective baseline while using fewer Chinese epochs; scaling capacity yields the strongest internal ARMOR-DP system. These results support a simple conclusion: multilingual weakness is a training-time state rather than a fixed property of a language. ARMOR-DP therefore separates diagnosis from intervention, allowing protection to follow the current bottleneck instead of a predetermined language label.

Auditing evaluation artifacts and data leakage in large language model systems for rare disease diagnosis

Jian Yang, Peilin Liu, Xiaohan Li, Songda Li, Yu Cao, Yuqing Feng — Tue, 27 Oct 2026 00:00:00 +0000

92% · Evaluation 92%, Agents 32%

Augmented large language model (LLM) systems have reported high benchmark accuracies in rare disease diagnosis, yet their real-world clinical utility remains unproven. A recent meta-analysis flagged augmented LLM diagnostic evaluations as having a high risk of bias due to suspected data leakage; however, the precise impact of these vulnerabilities on model inference has yet to be fully quantified. In this study, we present a comprehensive empirical audit to quantify evaluation artifacts across prevailing AI paradigms, using a repository of over 60,000 cases from diverse public and independent clinical cohorts. By reconstructing a state-of-the-art agentic framework, we demonstrate that candidate-space contamination artificially inflates performance: enforcing strict cross-dataset isolation resulted in a marked performance decline, with the cross-dataset mean Recall@1 decreasing from 54.3% (under conventional one-vs-rest evaluation) to 40.3% (Hedges' g = 0.94). Similar spatial overfitting in supervised classifiers caused diagnostic accuracy to drop from an internal mean of 42.9% to 16.7% across external cohorts (Cohen's d = 2.01). Furthermore, we quantify how LLMs infer diagnoses by exploiting retrospective therapeutic cues in clinical narratives, and how temporal misalignment in foundational knowledge bases leads to performance fluctuations. We conclude that prevailing evaluation protocols for augmented LLM systems inadvertently reflect data leakage rather than genuine medical reasoning. To address these systematic artifacts, we propose a rigorous evaluation framework - coupling strict data-isolation protocols with standardized benchmarking safeguards - to ensure future diagnostic AI for rare diseases is evaluated on actual clinical robustness.

SWE-Ladder: Learning to Localize, Reproduce, and Patch via Progressive Privileged Self-Distillation

Xudong Cai, Zhengxi Lu, Yuchen Yan, Jian Shao, Jun Xiao, Yueting Zhuang — Tue, 27 Oct 2026 00:00:00 +0000

92% · New model 92%, Agents 90%, Evaluation 31%

Resolving a repository-level issue typically requires a software-engineering (SWE) agent to localize the faulty code, reproduce the failure, and patch the code. Reinforcement learning trains such agents from a single test result on the final patch, while on-policy self-distillation (OPSD) adds token-level supervision from a teacher with privileged evidence, exposing the full reference patch at every step. Neither provides stage-specific guidance for localization or reproduction. We propose SWE-Ladder, which turns the verified repair of each training instance into stage-specific evidence: the changed files and lines for localization, the fail-to-pass tests for reproduction, and the diff for patching. A rule-based router assigns each step of an end-to-end rollout to a stage based on its tool call, a frozen teacher sees only the matching evidence, and the stages are distilled in repair order while GRPO optimizes the full trajectory. With Qwen3-4B-Instruct-2507 on SWE-bench Verified under the R2E-Gym scaffold, SWE-Ladder improves pass@1 from 9.6% to 16.4%, compared with 10.4% for GRPO and 12.4% for OPSD with the full patch. Under the OpenHands scaffold, it improves pass@1 from 13.0% to 17.6%. Reversing the stage order under R2E-Gym reduces pass@1 to 11.4%.

Intervention-Certified Amplification: A Validity Gate for $\sigma$-Weighted Post-Training

Leiyue Chen, Meng Wei, Xuefeng Hu, Wei Liu — Tue, 27 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 75%, Hallucination 32%

Hybrid post-training often mixes supervised fine-tuning (SFT) and reinforcement learning (RL), using training-side signals to set each sample's balance. Such signals characterize the anchor data and policy, but not whether the emphasized behavior helps in the target task and evaluation setting. This mismatch can reactivate and amplify pretrained behavior even after its literal form is removed from supervision. We demonstrate this with DIAA (Dual-Implicit Adversarial Amplification), a per-sample $\sigma$-weighted SFT/RL hybrid. With all literal Python fences removed from its anchor, DIAA emits unexecuted Tool-Integrated Reasoning (TIR) on $29%$ of MATH-500 prompts, three times vanilla GRPO. Paired interventions show $10.6$ percentage points of accuracy recovery when TIR is suppressed and comparable contributions from SFT imitation and $\sigma$-amplification; a student trained on a $0%$-TIR teacher anchor emits it at $45.8%$. Intervention-Certified Amplification (ICA) gates $\sigma$ with paired-rollout target evidence. Frozen and online ICA reduce TIR to $13.8%$ and $11.2%$, halve residual drag, improve accuracy from $47.5%$ to $48.7%$ and $49.3%$, and transfer to held-out GSM8K without retuning.

BriLLM-MetaPred: A Representation-Prediction Unity Language Model with Adaptive Self-Measurement

Junyuan Liu, Hongyue Lin, Jiale Hong, Yi Zhao, Hai Zhao — Tue, 27 Oct 2026 00:00:00 +0000

92% · New model 92%

We present BriLLM-MetaPred, the first engineering implementation of the MetaPred machine-learning framework established by a mathematical theory of AGI, instantiated for language modeling. The theory argues that MetaPred is both necessary and sufficient as a machine-learning primitive for general cognition. Taking this theoretical conclusion as a design premise, we report an implementation of the primitive for language modeling. A single continuously evolving dynamical state serves as both the context representation and the state from which predictions are read out. Continuous measurement of its own state reduces finite-depth inconsistency, maintains internal consistency, and provides an endogenous dynamical self-supervision signal in fixed-point form. The single dynamical state of MetaPred intrinsically couples training and inference. MetaPred therefore exhibits symmetry across almost all key elements of machine learning, an intuitive and observable property of the mathematical optimum over the space of all machine-learning paradigms. BriLLM-MetaPred is neither another recurrent language model nor an extension of BriLLM. It provides a bridge between two prior contributions: the construction and motivation of the MetaPred primitive provided by AGI theory, and the SiFu learning paradigm provided by BriLLM (Zhao et al., 2025). The latter's node-signal-flow design enables the primitive to run on available software and hardware accelerators. Our implementation of BriLLM-MetaPred achieves predictive performance comparable to a GPT baseline of the same size. This provides initial evidence for the engineering feasibility of the MetaPred framework established by the mathematical theory of AGI. MetaPred is the first machine-learning paradigm derived entirely from mathematical theory rather than assembled and incrementally improved through human design---and perhaps the last such paradigm.

PhysEvolve: Verifier-grounded Self-evolution for Physical Reasoning with Executable Worlds

Jinyuan Hu, Yilei Jiang, Chen Tang, Minghong Cai, Xinran Liu, Xiangyu Yue — Tue, 27 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 47%

Physical reasoning underpins how embodied systems understand and act in the physical world, yet vision-language models (VLMs) remain unreliable on questions about dynamics and physical causality. Progress is critically constrained by the scarcity of grounded counterfactual training data, as observational videos capture only the factual trajectory that occurred. The self-evolving paradigm offers a route beyond fixed datasets by enabling models to construct their own training problems. Extending this paradigm to physics, however, requires an external source of truth that can both validate model-proposed environments and verify alternative physical outcomes under intervention. We introduce \physevolve{}, a verifier-grounded self-evolving framework that trains VLMs for causal physical reasoning from model-proposed executable worlds and simulation-authored tasks. At each round, a shared VLM proposes executable world programs and evidence schemas; a deterministic physics engine freezes each valid world, simulates matched baseline and body-elided counterfactual branches, and synthesizes discrete, answer-bound visual tasks. Exact verifier feedback from task solving updates both the shared policy via GRPO and an adaptive curriculum over physical intents. Starting from Qwen3-VL-8B-Instruct, \physevolve{} achieves targeted gains across PhysBench, CausalPhys, and ContPhy without human intervention data, improving PhysBench Throwing by $8.8$ points and ContPhy Counterfactual Removal and Goal-Pit by $7.4$ and $24.4$ points. These gains concentrate on interaction-sensitive and altered-world questions, demonstrating zero-shot transfer of causal principles to unseen complex dynamics (cloth, fluid, soft bodies) and real-world video.

BellowsKV: One Latent for Many Layers, Expanded on Demand

Jiamu Zhang, Hanjie Chen — Tue, 27 Oct 2026 00:00:00 +0000

92% · Efficiency 92%

Large language models (LLMs) now serve as the backbone of agents that reason over many turns, call tools, and pursue long-horizon tasks. To avoid recomputation, a model stores the attention keys and values of all past tokens in a key-value (KV) cache. Each layer maintains its own cache, so KV-cache memory can become the main bottleneck for GPU serving. Most compression methods operate within a single layer, by quantizing the cache entries, reducing the channel dimension, or evicting less important tokens. All three lose substantial accuracy at high compression ratios. Cross-layer methods share one compressed cache across adjacent layers to go beyond this limit. However, these methods also degrade sharply at high compression ratios. We find that the KV-cache representations of adjacent layers are quite different, whereas the input hidden states of adjacent layers are highly similar. We also find that, under the same memory budget, reducing the dimension of this shared latent loses far more accuracy than quantizing it. Based on these findings, we propose BellowsKV, a training-free method that maps the hidden states of n adjacent layers to one shared latent per token. At inference time, each layer reads its own keys and values from this latent through a layer-specific projection. We keep the latent at full dimension and compress it in two other ways: BellowsKV-Quant stores the latent at low precision, which is nearly lossless because the latent has no dominant outlier channels; BellowsKV-Evict makes one shared eviction decision per block, because the layers rank token importance very similarly. Across Llama-3.1-8B, Mistral-7B and Qwen2.5-7B, BellowsKV-Quant retains about 99% of full-cache accuracy at 3.6x compression and 97-98% at 13-14x, where the strongest cross-layer method retains at most 84% at 16x. Code: https://github.com/MorrisZJ/BellowsKV

TUON: A Tensor-Aware Optimizer Utilizing Tensor-SVD and Polarization for Large Language Model Pretraining

Anda Tang, Xuanlin Yang, Junjie Wang, Zhouchen Lin — Tue, 27 Oct 2026 00:00:00 +0000

92% · New model 92%

Large scale language model~(LLM) pretraining is increasingly bottlenecked by optimizer efficiency, driving optimizer design from vector-aware methods such as AdamW to matrix-aware methods. Matrix-aware optimizers such as Muon exploit matrix structure but treat layers independently, neglecting cross-layer correlations. Tensor-aware optimizers are therefore needed, yet existing ones rely on a matricization-based tensor algebra and remain restricted to shallow stacking and specific modules. In this paper, we propose \textbf{TUON}, a novel optimizer built on a transform-based tensor algebra derived from the Tensor-SVD (T-SVD). Unlike methods that polarize a globally unfolded matrix, TUON mixes layers via a unitary transform $\boldsymbol{Q}$ and then polarizes each transformed slice, and inverts the transform to recover the update. This mix-then-separate strategy captures cross-layer correlations while preserving per-layer diversity, avoiding the structural dimension imbalance and cross-layer misalignment of unfolding. Thus it successfully scales to more layers and diverse modules beyond QKV. Theoretically, we prove that TUON performs steepest descent under a $\boldsymbol{Q}$-transformed tensor spectral norm, and identify its update as the tensor polar factor of the T-SVD. We further establish transform non-redundancy, norm and smoothness comparisons with Muon and TEON, and a convergence guarantee. Empirically, TUON improves consistently over Muon and TEON on GPT-124M/350M and Qwen-0.6B/4B, with ablations confirming the transformation mechanism, cross-layer structure beyond two-layer stacking, adaptability beyond QKV, and robustness to the polarization method.

Amplify, Locate, Estimate (ALE) on Rare-Failure Rates of Fine-Tuned LLMs

Patrick Reichherzer, Jonathan Evertz, Thorsten Holz — Tue, 27 Oct 2026 00:00:00 +0000

92% · Evaluation 92%, Efficiency 70%

Safety fine-tuning can make undesirable model behaviors sufficiently rare that measuring their frequency by direct sampling becomes prohibitively expensive. We introduce Amplify, Locate, Estimate (ALE), a method for estimating such rare failure probabilities efficiently. ALE uses the change in token log-probabilities induced by the safety fine-tuning to define an amplification direction, along which we increase the failure rate for cheap measurement. We develop a score-function estimator that uses the selected amplified batch to recover the local slope of the log failure rate from the token probabilities of the sampled failures, enabling extrapolation back to rare target rates. We also use the amplified batch to obtain an importance-sampling estimate. Across 41 evaluation targets, ALE's estimate-to-reference ratios have a median close to one, with 37 ALE estimates within a factor of three of reference values obtained from independent direct measurements. Ablation tests show that with a budget of up to $\sim$1000 labeled rollouts, in addition to a cheap unlabeled scan, ALE estimates failure probabilities for Qwen3-8B fine-tunes down to about $10^{-5}$, where direct measurement can require hundreds of thousands of rollouts, as the required sampling budget scales inversely with failure probability at fixed relative precision.

Scaling Qualitative Analysis with Large Language Models: Development and Validation of the Smart Coder Framework for Human-LLM Collaborative Coding

Tue, 27 Oct 2026 00:00:00 +0000

91% · Evaluation 91%

Current uses of large language models (LLMs) for qualitative coding often rely on ad hoc prompting with limited guidance for reliability and human oversight. To address these limitations, we propose the Smart Coder Framework, a three-phase approach structuring human-LLM collaborative coding through Instruction, Prompt Strategy, and Reliability Check. We evaluated the framework using 72 middle-school mathematics teacher interview transcripts and a human-consensus benchmark. A factorial design compared instructed versus uninstructed prompts, few-shot versus few-shot plus chain-of-thought prompting, and single-pass versus self-consistency coding. Twelve transcripts were used for prompt development and calibration, and 60 for the formal evaluation. In the results, instruction consistently improved agreement. Chain-of-thought further improved agreement under adequate instruction, and self-consistency produced smaller, theme-dependent gains but helped identify unstable decisions for human review. Performance also varied by transcript characteristics. Overall, the framework systematically structures and monitors LLM-assisted qualitative coding while preserving human authority over coding decisions.

Scoped Semantic Invariance as a Foundation for LLM Reliability

Rosario Uceda-Sosa, Maria Chang, Karthikeyan Natesan Ramamurthy, Moninder Singh — Tue, 27 Oct 2026 00:00:00 +0000

91% · Evaluation 91%, Hallucination 63%

LLMs often answer semantically equivalent queries inconsistently, even when the underlying facts are known. We formalize consistency as semantic invariance over equivalence classes of queries and show that invariance under arbitrary meaning-preserving transformations is undecidable in general. This motivates scoped semantic invariance, where model behavior is evaluated relative to bounded semantic domains. We operationalize this framework using knowledge graphs (KGs), which act as explicit semantic control structures defining local semantic constraints and grounding contexts. Across the MITRE D3FEND ontology and eight Wikidata- derived domains, we show that inconsistencies are structured rather than random, concentrating in localized hotspot subgraphs organized around recurring TREE, STAR, and CYCLE patterns that remain stable across models and domains. We further show that selective graph grounding improves correctness and cross-model consistency, while indiscriminate grounding introduces regressions. A layered grounding strategy achieves the strongest monotonicity profile, preserving correct knowledge while reducing inconsistency. Finally, we show that hotspot structures discovered from simple consistency probes transfer to more realistic compositional reasoning tasks.

LSTM-Style Looped Transformers with Shared KV Caches

Andreas Kirsch — Tue, 27 Oct 2026 00:00:00 +0000

91% · New model 91%, Efficiency 85%, Evaluation 39%

Recurrent transformers reuse weights to perform deep computation with fewer parameters. We study an LSTM-style transformer that also shares one key--value cache across its recurrent core. At width 1024, it uses 40.4% fewer parameters than an untied control at nearly matched forward FLOPs, and reduces persistent cache sets from fifteen to four. Parallel training supplies cache entries from a standard-attention pass, whereas continuation reads entries produced recursively using shared attention. We address this mismatch by aligning the two training passes' cache entries with a consistency loss during the final 10% of a fixed training-token budget. Across three model widths on FineWeb, this removes 81--89% of the continuation-loss penalty relative to full-cache evaluation. The improvement replicates on disjoint documents for two training seeds. Across four widths and 1.05--4.19B training tokens, the loss cost of weight tying is smallest at lower token budgets, suggesting an opportunity to use recurrent computation when data and model storage are limited.

FedSEAM: Federated Continual Learning of LLMs via Subspace Expert Agreement Merging

Fatema Siddika, Abhijit Chunduru, Zilinghan Li, Ali Jannesari, Ravi Madduri — Tue, 27 Oct 2026 00:00:00 +0000

91% · New model 91%, Evaluation 35%

Federated continual learning trains a model on a sequence of tasks across several clients without sharing their data while retaining knowledge of earlier tasks. Sparse mixture-of-experts models support this setting by adapting selected parameters on a frozen backbone. However, clients learn different expert content as tasks evolve, making it difficult to share knowledge across clients without disrupting task-specific and client-specific specialization. Matching tile coordinates alone does not establish whether their learned content is compatible. We present FedSEAM, a federated continual learning framework that organizes task knowledge into sparse experts and selectively shares their content across tasks and clients. On each client, task-aware tile selection allocates capacity to a new task, while a subspace-based compatibility test determines which existing expert tiles to reuse and where to introduce fresh capacity. On the server, Subspace Expert Agreement Merging (SEAM) compares client updates as subspaces and moves each client towards consensus only along agreed-upon directions, leaving the rest of its update unchanged. We identify two remaining sources of forgetting, routing drift and merge erasure, and address them with three protections, one of which exactly preserves tile components in a selected prior-overlap subspace at every merge. We evaluate FedSEAM on TRACE using Qwen3-4B and Llama-3.2-3B, with component-wise ablations and studies of data size to assess its balance between learning new tasks and retaining prior knowledge.

Eat Your Precision: Precision Objective Mitigates Reward Hacking in Rubric-based RL

Hyun Ryu, Jong-Heon Kim, Yiqing Xie, Seungone Kim, Carolyn Rose, Eunho Yang — Tue, 27 Oct 2026 00:00:00 +0000

91% · New model 91%, Hallucination 88%, Evaluation 52%

Rubric-based rewards enable effective reinforcement learning on non-verifiable tasks. However, because these rewards are typically defined as a weighted measure of rubric coverage, models are trained primarily to maximize a recall objective. As a result, RL-trained models produce responses that are more complete but also more verbose and contain more unnecessary or factually incorrect content, leading them to be less preferred than responses from the base model. To address this issue, we present the FORK reward, which jointly optimizes recall and a novel precision objective, which requires each unit of a model response to be both necessary and factually correct, as measured by rubric-driven necessity credits and a factual-correctness grader, respectively. Our experimental results show that FORK outperforms existing rubric-based RL methods and achieves Pareto-optimal performance that balances rubric coverage and overall response quality. Compared to the original recall-based reward, FORK substantially improves precision and reduces response length while achieving similar rubric coverage, which leads to significantly higher win rate compared to the base policy (e.g., from 10.6% to 68.0% on RubricHub-Medical and from 22.1% to 74.1% on HealthBench).

Implicit Beliefs as Scalable Process Rewards for Strategic Persuasion

Minqian Liu, Yuexi Shen, Heajun An, Qi Zhang, Pamela J. Wisniewski, Sang Won Lee — Tue, 27 Oct 2026 00:00:00 +0000

91% · New model 91%, Evaluation 89%

Training large language models (LLMs) for multi-turn strategic persuasion faces two coupled obstacles: (i) realistic simulation of psychologically grounded persuadees, and (ii) dense, faithful supervision for credit assignment in strategic dialogues. Outcome-only rewards are sparse and uninformative about which conversational moves drove success, while LLM-judge process rewards are expensive and decoupled from the underlying interaction dynamics. We address both challenges with a unified framework. First, we introduce PersuArena, a persuasion simulation environment that produces realistic, non-monotonic trajectories. Second, building directly on this environment, we propose Process Rewards via Online Belief Elicitation for Relative Policy Optimization (Probe-RPO), an online RL framework that converts the persuadee's internal state into dense process rewards. At each turn, three introspective probes are read from the frozen persuadee's logits, yielding a multi-dimensional reward signal that is faithfully aligned with persuasion dynamics. With extensive experiments on PersuArena spanning healthcare, financial advising, workplace management, education, and personal relationships, we demonstrate that our Probe-RPO achieves the full acceptance rate of 52.5%, compared with the 35.8% for the strongest GRPO ablation and 25.8% for the best off-the-shelf LLM. We also conduct comprehensive analyses on the generalization across persuadee models and the persuasion dynamics across conversations. Our work offers a promising paradigm that can provide effective process supervision for strategic dialogue.

ReGE: Stronger Agents through Experience Synthesis

Jiarong Liang, Yubo Wang, Zijie Liu, Yuxuan Zhang, Ping Nie, Wenhu Chen — Tue, 27 Oct 2026 00:00:00 +0000

91% · Agents 91%, Efficiency 87%, Evaluation 38%

Repeated attempts at interactive tasks uncover facts, useful operations, and failures that can inform subsequent execution. We introduce ReGE, a framework for turning this recorded experience into guidance for a fresh attempt. Its main implementation, ReGE-Syn, uses a single coach call to jointly synthesize bounded execution records and available artifacts into self-contained guidance, connecting task-specific findings with proposed actions and relevant pitfalls. A student starts from the task's initial environment state and adapts this guidance to current observations. Source generation, synthesis, and student execution use the same base model, without parameter updates or a stronger teacher. Across four benchmarks, ReGE improves accuracy by an average of 9.9 percentage points for GPT and 8.8 percentage points for Claude compared with unguided execution. On matched BrowseComp-Plus subsets, ReGE-Syn reduces guidance-construction tokens by 98.2\% on GPT and 95.5\% on Claude relative to adapted AggAgent, with observed success. Combined construction and execution tokens decrease by 15.7\% and 35.4\%, respectively, counting input plus output and excluding source generation. BrowseComp-Plus ablations further associate access to process information with fewer student steps and show that joint synthesis produces substantially shorter guidance than Direct history or Independent summaries, with observed success. These results demonstrate how agents can use their own recorded experience to improve subsequent task completion, and suggest a low-cost way to consolidate the execution records of parallel agent systems such as agent swarms.

AndroidFlux: Evaluating and Reward Modeling Failure Recovery Capabilities in Mobile Use Multimodal Agents

Gyubeum Lim, Yemo Koo, Ghunwoo Jang, Zsolt Kira — Tue, 27 Oct 2026 00:00:00 +0000

91% · Evaluation 91%, Agents 53%, New model 38%

Mobile-use multimodal agents are typically evaluated by task success from a clean initial state, which does not isolate their ability to recover from intermediate errors. Because mobile environments change dynamically, a general mobile-use agent must complete tasks robustly even from unexpected states and in unseen environments. This requires knowing when and where an agent makes errors, and whether it can recover from errors and subsequent states. To evaluate this capability, we introduce AndroidFlux, a failure-recovery benchmark built on AndroidWorld. AndroidFlux restores recorded trajectories to controlled checkpoints before, at, and after the first recoverable error, so that different agents continue from the same device state and history. On these checkpoints, a complementary reward-modeling benchmark evaluates recovery-action preferences using downstream rollout outcomes. We further release an offline reinforcement learning (RL) dataset that improves both error recovery and overall task success. Our evaluations show that recovery performance only partially correlates with clean-start success, and that explicit reasoning helps only models trained with GUI-domain specific RL.

HARP-LoRA: Hierarchy-Conditioned Rank Projection for Low-Rank Adaptation in Scientific Table Understanding

Tue, 27 Oct 2026 00:00:00 +0000

91% · New model 91%, Efficiency 53%, Evaluation 45%

Tabular data is widely used in scientific applications, where complex layouts, including multi-level headers, merged cells, captions, and footnotes, make information extraction difficult. This is particularly important in pharmacokinetics (PK), where values must be correctly associated with the drug, species, dose, administration route, unit, and experimental condition. Parameter-efficient fine-tuning methods such as Low-Rank Adaptation (LoRA) reduce adaptation cost but do not explicitly incorporate relationships between table values and surrounding headers. We introduce HARP-LoRA, a hierarchy-conditioned rank projection method for low-rank adaptation in scientific table understanding. HARP-LoRA extends projected LoRA by constructing token-specific intrinsic representations from the row- and column-header hierarchies of each data token. These representations condition the low-rank down-projection, incorporating hierarchical context while keeping the pretrained model and random projection tensors frozen. We evaluate HARP-LoRA on 950 manually annotated PK tables from 587 scientific articles and on WikiTableQuestions (WikiTQ) using Llama~3.2 3B, Llama~3.1 8B, and Gemma~3 27B. With Llama~3.1 8B, HARP-LoRA achieves a PK F1-score of 57.64\% and WikiTQ accuracy of 53.15\%, improving over LoRA by 1.42 and 6.56 points, respectively. With Gemma~3 27B, it achieves the best PK F1-score of 70.63\% and WikiTQ accuracy of 56.87\%. These results use only 172K, 197K, and 547K trainable parameters, far fewer than LoRA and TableLoRA. Overall, conditioning lightweight adapters on table-header hierarchies improves extraction from scientific tables and reasoning over flatter general-domain tables while maintaining parameter efficiency. Code is available at: https://github.com/hosseinsholehrasa/HARP-LoRA

Semantic Atoms for Scalable Text Analysis in Data Analytics in the Era of LLM-Based Agents

Jinxiang Xie, Zihao Li, Rui Ding, Bartuer Zhou, Shi Han, Dongmei Zhang — Tue, 27 Oct 2026 00:00:00 +0000

91% · Evaluation 91%, Efficiency 84%, Agents 33%

Much of the information needed for data analysis is buried in text columns: an analyst may infer customer frustration from complaint text and use it to explain resolution time. We call workloads in which processing text columns is central to subsequent data analysis TADA (\emph{Text Analysis in Data Analytics}). In an agentic setting, the user's query dynamically specifies this processing, and answering it interleaves LLM-based semantic steps with exact SQL or Python operations. This mixed execution raises two design questions: which work should be neural and which symbolic, and which semantic outputs later steps can reuse. We address them for agent-planned TADA workflows with two complementary, query-driven semantic atoms. Each atom takes a sequence of text items and a natural-language query specifying the semantic task: \textbf{Tagging} returns one string per input item, while \textbf{Categorization} returns $M$ category objects over $N$ input items. Tagging results are materialized as a row-aligned column and Categorization results as a category table; SQL or Python then performs exact computation over this exposed state. We realize this interface in TADA-Agent, whose atoms run inside generated Python programs and internally handle inputs beyond a reliable context window. We evaluate TADA-Agent on TABench, a benchmark of 40 complete analysis requests over spreadsheet tables spanning workflow composition, input--output structure, and literature-grounded analytical scenarios. Against a semantic-operator baseline and relative to a held-out TADA reference run, TADA-Agent delivers comparable row-level semantic quality, has higher complete-artifact preservation estimates, and uses over an order of magnitude fewer semantic-model requests under the evaluated native execution paths.

Decision Model Factory: Building Specialized Sovereign Decision Models

Yuma Ichikawa — Tue, 27 Oct 2026 00:00:00 +0000

91% · Efficiency 91%, New model 69%, Evaluation 67%

Decision needs vary with each application, language, and deployment environment. The Decision Model Factory turns those needs into specialized, sovereign models on demand. A brief recipe guides the training, calibration, verification, and evaluation of a selected backbone, delivering a deployable decision model under the operator's control. The same workflow creates a new specialist whenever the requirements change. Across Japanese decision tasks, compact deployment, and long-context passage ranking, five factory-built models improve upon their backbones; three surpass the strongest results in a fixed public Japanese benchmark comparison. A dedicated runner computes only permitted-answer probabilities and reuses shared context, more than doubling serving speed on matched multi-question workloads with essentially unchanged decision scores. Together, the factory and runner make fast, task-specific decision models practical to build and deploy whenever needed.

Diagnosing and Mitigating Long-Session Degradation in Full-Duplex Speech Models

Junho Kim, CheHyun Lee, Donghun Kang, Youngho Cho, Dongwook Lee, Heeseung Kim — Tue, 27 Oct 2026 00:00:00 +0000

91% · Evaluation 91%, New model 88%, Efficiency 82%

End-to-end full-duplex spoken dialogue models enable natural, real-time spoken interactions with interruptions, backchannels, and overlapping speech. Prior evaluations have assessed diverse interaction capabilities and identified some degradation patterns across conversational turns. However, how model behavior evolves over extended sessions remains insufficiently understood. In this work, we address this gap by systematically characterizing long-session degradation and developing efficient methods to mitigate it. We introduce a closed-loop evaluation framework for long-session stability and apply it to several publicly available full-duplex speech models, revealing degradation over time in responsiveness, spoken content quality, and voice consistency. To mitigate this degradation, we propose a self-distillation-based training framework for lightweight LoRA adaptation that improves long-session stability with a small amount of additional training. We further introduce a two-resolution context (TRC) architecture that preserves recent interactions in full detail while compressing older interaction history into key-value representations. We evaluate our methods on three models with distinct architectures, assessing long-session stability, information retention, and computational efficiency. Across all three models, we show that LoRA adaptation mitigates performance degradation over ten-minute sessions, while adding TRC extends these gains beyond the training session length and reduces KV-cache growth by more than 90%.

Rethinking Evaluation for Principled Generalization Using a Kripkean Challenge to Meaning as Use

Niyar R Barman, Omar Naim, Nicholas Asher — Tue, 27 Oct 2026 00:00:00 +0000

91% · Evaluation 91%

Large language models operationalize an intuitive, view of meaning as use. We argue that it faces a rule-following challenge. Using Kripke's interpretation of Wittgenstein, we argue that finite histories of use do not determine a unique rule for future extension. This reframes the debate about OOD generalization but also about NLP evaluation; benchmarks that measure performance on fixed use cases or local robustness, we argue, do not by themselves establish semantic competence. To evaluate whether models preserve rules and concepts under principled domain extension, we propose to evaluate two capacities, central to this ability: abstraction and recursion; and we offer an experimental paradigm for testing both.

LoopFold: Parallelizing Recurrent Depth in Looped Transformers

Feichen Song, Boyi Zeng, Yiqin Hao, Shixiang Song, He Li, Yizhou Yang — Tue, 27 Oct 2026 00:00:00 +0000

91% · Efficiency 91%, New model 88%

Looped Transformers increase computation per token by repeatedly applying a shared backbone, offering a way to improve language modeling and downstream task performance at a fixed backbone parameter count. However, each loop depends on the preceding loop, so increasing recurrent depth also increases decoding latency. Existing approaches include Parallel Loop Transformers, which reduce latency by overlapping loop computation across token positions, but yield rapidly diminishing marginal accuracy gains at larger depths. To address this challenge, we introduce LoopFold, which folds computation at multiple recurrent depths into parallel batches for the same token prefix. A lightweight predictor estimates inputs to later loops so they can start before earlier loops finish. A corrector then refines the outputs of these later loops using newly available results from earlier loops. All components are trained jointly using only the final next-token prediction loss. In experiments with 1.4B models, deeper LoopFold configurations achieve progressively higher average zero-shot accuracy across eight tasks, rising from 53.41% to 56.38%. Vanilla looped models and PLT show limited gains over the same depth range. Compared with a nine-loop vanilla model, LoopFold uses five sequential stages and additional parallel computation to reduce decoding latency by 46%.

Entity-Aware Evaluation of OCR Systems for Historical Document Transcription

Tue, 27 Oct 2026 00:00:00 +0000

90% · Evaluation 90%

Optical Character Recognition (OCR) is essential for digitizing historical archives, yet standard evaluation metrics such as Character Error Rate (CER) do not necessarily reflect whether meaningful information, such as names, dates, and identifiers, is preserved. In this work, we investigate entitylevel analysis of OCR errors as a complementary perspective for evaluating OCR quality in historical documents. Using a publicly available benchmark of 175 documents from the Berrutti Archive, augmented with 5,209 manually annotated entity mentions that will be released upon acceptance, we compare sequence-recognition OCR systems, OCR-specialized vision-language models, and general-purpose vision-language models. Beyond global transcription accuracy, we analyze entity preservation and recoverability under varying character-level edit tolerances. Our experiments show that OCR rankings depend on the evaluation criterion: while Qwen 3.6 achieves the highest entity coverage across all tested tolerances, several systems with lower exact-match coverage recover substantially when small errors are tolerated, suggesting that many of their errors correspond to localized character-level deviations rather than fundamentally incorrect entity transcriptions. Conversely, some systems remain unable to recover many entities even under relaxed matching. These findings show that OCR quality cannot be fully characterized by aggregate transcription metrics alone and highlight the importance of evaluating how OCR errors affect the preservation and recoverability of meaningful information.

From Document Layout to Causal Topology: An End-to-End Architecture for Temporal Causal Reasoning

Zhifeng Yang, Wei Luo, Yongjian Chen, Shiqi Ou — Tue, 27 Oct 2026 00:00:00 +0000

90% · New model 90%, Evaluation 86%

Large language models (LLMs) generally lack deep causal reasoning capabilities in temporal reasoning. Inspired by document layout reordering, this paper proposes a paradigm shift from discrete narratives to causal topologies. It constructs two architectures: a training-free neuro-symbolic framework, temporal causal reasoning agent (TCR-Agent), which achieves interpretable causal deduction and counterfactual truncation through external causal graph instantiation and a symbolic intervention engine; and an end-to-end model, temporal causal reasoning former (TCR-Former), which internalizes these mechanisms into the Transformer attention space via a Causal-Topological Attention Mask and temporal span biases. Experiments on the TempoCausal benchmark demonstrate that TCR-Agent comprehensively surpasses existing training-free baselines. At the same time, TCR-Former achieves an accuracy 16.5% higher than that of state-of-the-art (SOTA) methods, outperforming large-scale models such as GPT-5.4 and DeepSeek-V3.2 with only approximately 8.6B parameters. Further analysis reveals a structure-reasoning disconnection phenomenon in existing methods: external symbolic constraints and chain-of-thought reasoning can improve reasoning steps, yet fail to enhance deep causal deductive capability. By internalizing temporal causal reasoning capabilities into the model, TCR-Former effectively bridges this gap, providing an efficient and reliable end-to-end foundation for complex reasoning scenarios.

STEP: Soft Trajectory Estimation of Prediction Correctness via Per-Step Probability Scoring

Chenyang Yu, Zifan Zeng, Qunli Zhang, Feng Liu — Tue, 27 Oct 2026 00:00:00 +0000

90% · Evaluation 90%, Hallucination 72%

Estimating the correctness of large language model (LLM) predictions without access to ground truth is critical for reliable deployment in high-stakes settings. Existing approaches, such as entropy and perplexity, operate solely on final output statistics, discarding the rich information encoded in intermediate reasoning steps. To address this limitation, we propose STEP (Soft Trajectory Estimation of Prediction correctness), a training-free framework that scores each reasoning step via lightweight forward passes to obtain a soft probability distribution qi over candidate answers, from which we derive a set of soft trajectory features capturing per-step confidence and its progression toward the final answer. Extensive experiments on various benchmarks demonstrate consistent and substantial improvements, particularly in AUROC, over output-based baselines and hard-indicator trajectory methods. Notably, Temporally-Weighted Consistency, which assigns higher importance to later reasoning steps where answer distributions have stabilized, contributes the largest gains, highlighting the importance of temporal structure in reasoning trajectories. These results further validate the advantage of preserving continuous probability signals over binary discretization.

CSI-TextBench: A Dataset and Benchmark for Language-Grounded Ambient Sensing Perception

Tue, 27 Oct 2026 00:00:00 +0000

90% · Evaluation 90%, New model 30%

Indoor context deciphered by ambient sensing underpins many applications, from assisted living to smart buildings, yet remains weakly connected to the language-aligned paradigm of modern multimodal systems. Through billions of IoT devices, ambient WiFi Channel State Information (CSI) captures the indoor context of human activity and environmental dynamics passively via ubiquitous, always-on infrastructure. However, unlike vision or audio, CSI lacks naturally paired text like captions or transcripts, limiting its integration with language-driven AI systems. We present CSI-TextBench, a large-scale dataset and benchmark for aligning CSI data with natural language, unifying 3 sensing benchmarks into 285,790 samples across 7 tasks and pairing them with 2,291 hierarchically structured, label-derived descriptions via many-to-many assignment to construct 2.0 million CSI-text training tuples. Benchmarking 5 alignment methods from 4 paradigm families, we show that contrastive CSI-text alignment matches supervised classifiers without class labels, adapts to unseen classes with few samples, and responds to open-vocabulary natural-language prompts. A single unified model supports compositional scene understanding, answering joint queries over activity, proximity, and identity, achieving 86.9% top-1 accuracy on 140 composite scenes, outperforming task-specific classifier ensembles. CSI-TextBench enables ubiquitous ambient sensing as a language-indexed perception layer, enabling indoor contextual awareness in language-driven AI systems.

Entity Matching using Decision Models

Aaron Steiner, Ralph Peeters, Christian Bizer — Tue, 27 Oct 2026 00:00:00 +0000

90% · Efficiency 90%, Evaluation 42%

Entity matching is the task of deciding whether two records describe the same real-world entity. Current entity matching methods rely on pre-trained language models (PLMs) which require task specific fine-tuning or generative large language models (LLMs) which do not require task-specific fine-tuning but are slow and expensive at matching large record sets. Decision models are a new class of LLM-based models that return a probability for each predefined answer option rather than generate text. This paper investigates decision models as an alternative to generative LLMs and fine-tuned PLMs for entity matching. We evaluate six decision models in a zero-shot setting across eight benchmarks and compare their performance with that of the generative LLM GPT-6 Luna and the PLM-based matcher Ditto. Our experiments show that, without task-specific training data, the four strongest decision models achieve mean F1 scores close to that of GPT-6 Luna, with the best model slightly surpassing it. This model also outperforms Ditto on six of the eight benchmarks. All decision models are faster than GPT-6 Luna, and most are less expensive. For three of the four strongest decision models, the predicted match probabilities are well calibrated without requiring any task-specific recalibration. A model cascade that routes the least confident record pairs to an ensemble of generative LLMs improves the mean F1 scores of the four strongest decision models by 1.9-4.0 percentage points. Using the best decision model and its best-performing prompt for each benchmark, the cascade surpasses the highest published F1 scores from literature on four of the eight benchmarks.

Infinity Machine: a self-improving AI agent for research

Harish Santhanalakshmi Ganesan — Tue, 27 Oct 2026 00:00:00 +0000

90% · Agents 90%, Evaluation 40%

We present the Infinity Machine, a self-improving AI agent that we use for research on problems whose answers aprogram or data can check. Each candidate solution is produced by a full tool-using agent that works for up to an hour in a fresh sandbox; proposers see only a public score, and a private split decides what counts. The harness itself(prompts, tools, skills, playbook, meta agent) is an editable genome, changed only through a kernel-enforced gate with held-out and cost checks. An autonomous mode goes from a plain-text goal to a verified report with no human step: an agent picks an open problem and writes its verifier, the kernel calibrates it, an adversarial agent audits it, and an agent that never sees the verifier re-checks the result. Across fourteen problems in six days, for about $180 of model usage, the system produced four results beyond the published state of the art on catalogued open problems in quantum information: a certified depolarizing-channel threshold point p = 0.064966, a qutrit positivity frontier of 0.2910672, a 44-gate depth-6 Golay state preparation, and a stabilizer-rank bound chi(T^7) <= 10. It also cut SABRE qubit-routing overhead from 68,559 to 12,087 added gates. In hardware, it cut the energy per MAC of an open 8x8 INT8 systolic tile by 77.7% through a full open RTL-to-GDS flow, bit-exact. Three fully autonomous runs ended in honestties with public records; one audit caught a stale record, and one agent-written score pointing the wrong way went unnoticed. We report the failures as carefully: the one self-derived harness generation was rejected on cost, proposers gamed lenient scorers four times, and operator changes were decisive for several results.

From Self-Confidence to Statistical Accuracy: Scaling Human-in-the-Loop Evaluation with Direct and LLM-Assisted Auditing

Tue, 27 Oct 2026 00:00:00 +0000

90% · Evaluation 90%

For large language models (LLMs) and multimodal large language models (MLLMs), model-reported self-confidence is often miscalibrated, yet it may still be useful for efficient evaluation. At the same time, LLM-as-a-judge offers a scalable alternative to direct human auditing, but its practical benefit depends on calibration quality and inference cost. This paper develops a unified statistical framework for both evaluation routes under alternative sampling strategies. For direct auditing, confidence-stratified sampling is used to recover population accuracy from partially verified outputs without requiring calibrated confidence. Finite-sample uncertainty is characterized, and the analysis shows how highly uneven confidence strata affect audit efficiency and label allocation. For judge-assisted evaluation, the judge is modeled through sensitivity and false-positive rate, yielding an error-corrected estimator with calibration uncertainty. The two routes are then compared through human-label and total-cost efficiency boundaries. Experiments on MMLU-Pro and MMMU-Pro show that miscalibrated self-confidence can still support statistically valid stratified auditing. Under highly imbalanced confidence distributions, proportional allocation substantially reduces estimation error relative to equal-per-stratum allocation. For judge-assisted evaluation, only a small subset of target--verifier pairs satisfies the idealized human-label efficiency condition, and automated inference costs can further limit potential savings. These findings clarify the statistical and economic conditions under which confidence-guided auditing and automated verification are useful.

★ MisconceptDialogue: Evaluation Criteria Shape Measured Success in LLM Math Tutoring

Woojin Lee, Yoonkyu Woo, Jinah Kim, Hyeok-Min Gwon, Jin-Xia Huang — Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Large language models are increasingly used as math tutors, raising two questions: can a single LLM detect a student's error, diagnose the underlying misconception, and remediate it through multi-turn dialogue - and can we trust the numbers that report how well it does? We address both with MISCONCEPTDIALOGUE, a fully automated framework that runs the full pipeline on five open-weight LLM tutors (7B-14B) interacting with LLM-simulated students over 5,000+ dialogues - a deliberately simulated protocol (paired conditions, a wrong-diagnosis control unsafe with real learners, scale) whose results do not establish real-student learning. On measurement, our headline finding is uncomfortable for the field: benchmark-style exact-match scoring common in standard math NLP - a lower-bound stress test - reports 42.1% measured Remediation Success Rate on identical dialogues where a human-calibrated LLM-as-judge reports 81.3%, a 39.2 percentage-point spread larger than the combined effects of tutor model, teaching strategy, and supplied diagnostic information. This is not a single-judge artifact: four independent LLM judges and 200 human annotations converge to within a few points (kappa up to 0.91). On capability, the verdict is equally striking: the score barely changes whether we tell the tutor the student's actual misconception or a deliberately swapped one - under the tested diagnosis-presentation format, these tutors show little evidence of reliably conditioning on the diagnostic information they are handed. For LLM tutoring evaluation to mature, multi-criterion scoring with multi-judge convergence and human anchoring should become standard practice - and the measurement question deserves as much scrutiny as the capability question it claims to answer.

★ The Schema Coherence Inventory: Measuring Whether an Assigned Belief Organizes a Language Model's Cognition

Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Persona-conditioned language models are used as simulated patients and affective agents, but there is no validated way to ask whether a belief assigned to such a model organizes its behaviour or whether the model has merely adopted a mood. We present the Schema Coherence Inventory (SCI): 96 forced-choice items at eight levels of Beck's cognitive model, each offering six replies keyed to three negative self-schemas and their positive counterparts, scored by length-normalized log-likelihood so that nothing is generated or parsed. Affect alone can move a model onto a default belief, so the discriminant criterion is target specificity: a condition built for one belief must select that belief more often than conditions built for the other beliefs of its pole. Six validation studies over 107 conditions on five backbones (1.7B-14B parameters) support this reading. Nine clinically trained raters recover the keyed belief at 79.5% (chance 16.7%). Belief adapters are target-specific on every backbone (+17.7 points; 30/30 belief cells positive), and a condition's SCI profile identifies its own free generations in 24/30 cases against 5 expected by chance. A 48-item short form correlates r = 0.73 with the other 48 items. Applied to six published patient simulators, none carries its formulation's core belief better than a one-sentence statement of it. The validation route, which separates a construct from the affect that accompanies it and checks model judges against clinicians, offers a template for affective measurement in language models; we outline its extension across task types, raters, sites and languages. We release the items, scoring code and short form.

★ SAGE: Sense-Aligned Multimodal Grounding Evaluation with Gloss-Augmented Reasoning

Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

This paper introduces Sense-Aligned Grounding Evaluation (SAGE), a unified 10-candidate image-selection benchmark that reformulates VWSD, VerSe, and WikiDiverse into a shared term-centric protocol covering noun-, verb-, and entity-centered grounding. SAGE includes confusable hard negatives and a context-swapped VerSe diagnostic to probe target-term grounding beyond direct caption--image compatibility. We further propose Gloss-Augmented Reasoning (GAR), a training-free framework combining definition grounding with constraint-guided semantic verification. Experiments show that strong VWSD performance does not transfer uniformly across grounding settings. With Gemma3-12B, definition augmentation improves all three subsets, while constraint-guided reasoning improves VWSD and yields particularly strong gains on VerSe, but is less effective on WikiDiverse. Combining the two signals achieves the best macro-average performance, improving Hits@1 from 65.90 to 71.22. These results show that SAGE supports both unified evaluation and fine-grained analysis of explicit semantic reasoning across grounding settings.

★ The Helpful Agent Problem: When Emergent Cooperation Among LLM Agents Bypasses Intended Workflows

Moumita Choudhury, Amit Sarker, Shlomo Zilberstein — Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

LLM agents are increasingly deployed in collaborative settings where they pursue separate tasks while sharing resources and operating under review or oversight processes. Such settings create safety risks not only when agents are malicious, but also when they are simply trying to help. Prior work has largely studied adversarial or explicitly collusive behavior; we show that benign agents pursuing separate objectives can spontaneously cooperate in ways that bypass intended workflows. We call this the \emph{helpful agent problem}. We study long-horizon shared-resource environments in which agents sometimes depend on resources owned by their peers and cross-owner changes are normally mediated by a reviewer. When review becomes unavailable, agents begin making changes on one another's behalf despite having no shared reward or explicit incentive to cooperate. Crucially, these actions remain within each agent's local permissions while circumventing the system's intended oversight structure. We introduce \textsc{FavorBench}, a controlled benchmark with measurable downstream side effects, and adapt \textsc{CooperBench} to test the phenomenon in realistic repositories. Across two benchmarks and four models, we find that out-of-workflow favors emerge frequently. Further, in \textsc{FavorBench}, favors from at least three models cause significant harm to parties outside the interaction. These results reveal a distinct multi-agent safety challenge: cooperation itself can create pathways around system-level safeguards.

★ DraMaze: A Multi-Agent Pipeline with Future-Aware Memory for Diverse Microdrama Script Generation

Siyang He, Qiqi Wang, Fan Liu, Yuncheng Huang, Pengfei Xu, Xi Chen — Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

The rapid growth of microdramas has increased demand for high-quality script production. Long-form microdrama script generation with LLMs requires both diverse plot progression and long-range narrative coherence. Our observations reveal that candidate continuations increasingly concentrate on fewer narrative mechanisms as generation progresses, while conventional retrieval may reinforce trope reuse and overlook historical information essential to future plots. To address these challenges, we propose DraMaze, a multi-agent pipeline that explores diverse causal paths through trope memory-guided evolutionary tree search and mutation. It further constructs a Backward Narrative Dependency Tree to organize, retain, and retrieve historical memories based on future narrative goals. We also introduce MiDramaBench, a bilingual benchmark for evaluating script logic and creative quality across different script lengths, together with causal-pattern metrics for narrative repetition and diversity. Experiments show consistent improvements across three backbone LLMs. For 60-episode scripts of up to 60k words, DraMaze reduces causal-pattern repetition by 32.1% and achieves a Logic score of 82.5, outperforming the best-performing baseline by 8.5 points.

★ BARC-Bench: Benign Asks, Risky Consequences

Kangkang Li, Kelvin Luu, Lily Ma, Tu Trinh, Ying Liu, Mohamed Elfeki — Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 64%

Task instructions from users for agents are inherently underspecified, especially regarding safety boundaries--creating behavioral risks. To study these risks, we introduce BARC-Bench, a collection of 110 tasks across nine simulated workspace environments where benign requests leave openings for unsafe actions. In this benchmark, we evaluate agent behavior against safety boundaries established by workspace records, implicit environmental context, and unstated social norms. For example, an agent asked to schedule off-boarding discussions must arrange the meeting without prematurely revealing an employee's resignation through a shared calendar invitation. Evaluating four frontier model configurations across five attempts per task reveals that models are not consistently safe: the share of tasks completed safely on all five attempts ranges from 9.1% to 72.7%, with even the strongest configuration producing a safety violation on 27.3% of tasks. These results highlight an asymmetric operational risk: while models frequently complete complex workflows, their inability to consistently respect unstated boundaries presents a substantial barrier to safe autonomous delegation.

★ PersonaScienceBench: Evaluating Scientific Agents with Diverse Simulated Researchers

Siyang Wu, Weizhi Zhang, Chen Yang, Esther Xu Fei, Jorin Overwiening, Xu Pan — Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 40%, Agents 38%

Scientific agents collaborate with researchers whose expertise and preferences differ, but benchmarks usually hold the collaborator fixed. We introduce PersonaScienceBench: 200 tasks from five benchmarks across four scientific domains. Simulated researchers provide context and feedback according to persona profiles, while task objectives and grading remain fixed. Agents encounter researcher preferences only through conversation. Across eighteen modelconfigurations, task performance, fact recovery, and self-repetition reveal different strengths. Detailed GPT 6 Astra analyses show distinct agent responses across researcher profiles; interaction traces illustrate targeted clar ification and workflow revision. These behaviors can differ even when final scores match. PersonaScienceBench therefore evaluates how agents collaborate with diverse researchers alongside whether they complete the task.

★ Tampered ML: Evaluating LLM Judgment of Machine Learning Research

Soroush Ebadian, Parand A. Alamdari, Joe Benton, Hailey Joren — Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

One approach to accelerating progress on AI alignment is to automate parts of alignment research itself. Doing so requires systems that can perform reliably on hard-to-supervise research tasks, including identifying flaws in research proposals and evaluating whether a proposed experiment supports its intended claim. To evaluate this capability, we introduce Tampered ML, a benchmark of 1,858 item-pairs, each consisting of an experiment proposal reconstructed from a published machine learning paper and a minimally edited counterfactual that introduces a targeted methodological flaw. Models must identify the more methodologically sound proposal in both presentation orders. A human validation study on a stratified sample estimates that 86% of items are valid (95% CI: 81-91%). Frontier models perform above chance but remain far from reliable (with Opus 5 achieving 77.9% accuracy). They are highly sensitive to the order of presentation and favor flawed designs that appear more specific, transparent, or conventional.

★ SWE-Shift: Evaluating Coding Agents in Evolving Workspaces

Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 80%

Repository-level coding benchmarks largely assume that tasks and environments remain unchanged throughout execution. In software development, code, dependencies, runtime state, and requirements can change while an agent is still working. We introduce SWE-Shift, a benchmark for evaluating coding agents under such mid-execution changes. From the same saved history and workspace, a Clean continuation proceeds without intervention, while a matched Shifted continuation receives a task-relevant update. Success requires completing the current task, preserving still-valid work, and satisfying the applicable state and evidence checks. The benchmark contains 120 scenarios across 12 change families, each evaluated at nine scheduled intervention positions. We evaluate 17 coding-agent systems, all of which have lower Shifted than Clean success. GPT-6 Astra drops from 67.35% to 24.79%, with 20.34% paired success. Tool/session changes and stale evidence yield zero paired success across all systems. Late-position paired success is lower for some systems and higher for others. These results show that success on the unchanged task can coexist with failures to preserve work, recover state, or refresh evidence after an update.

★ DMAIC-Bench: From Self-Reflection to Self-Restraint for Agents in Multi-Stage Analytical Tasks

Shan Peng, Shuaiqi Cheng, Miao Peng, Chenyi Zi, Jia Li — Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 52%

Multi-stage analytical tasks require LLM agents to use intermediate evidence to decide whether to continue, redirect or stop. Self-reflection, in which an agent critiques its output, is widely studied, yet a critique does not show whether evidence changed the next action. We call that capability self-restraint. Existing agent benchmarks largely measure task completion or output quality and leave it untested. DMAIC-Bench evaluates it with 97 seeded tasks across 15 industrial cases in four sectors, built around Lean Six Sigma's explicit statistical prerequisites. Paired counterfactual variants require opposite actions at the same workflow position, distinguishing warranted restraint from indiscriminate stopping. We assess execution traces with a private scorer and final claims through independent expert review. Across 13 models from six providers, execution-level routing accuracy averages 92% on matched GO decisions and 19% on NO-GO decisions, counting any forbidden downstream computation as a route error. Execution traces show agents reporting correct diagnostics before performing analyses that those results disallow. Separately, blinded expert review of the 11 audited agents estimates that reports act on invalid evidence at 47% of reached NO-GO decisions, with an interval from 31% to 61%. Explicit stopping permission helps some models without consistently restoring correct decisions at both GO and NO-GO forks. Sixteen life-sciences tasks in bulk ribonucleic acid sequencing (RNA-seq) and case-control genome-wide association studies (GWAS) show that the same framework can evaluate evidence-dependent decisions beyond industrial workflows. Project code and data are available at: https://github.com/psptpx/DMAIC-Bench-public.

★ Why Full-Duplex Speech Models Collapse in Long Conversations

Yeonju Kim, Bella Godiva, Se Jin Park, Yong Man Ro — Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Full-duplex speech models listen and speak at the same time, which makes them promising for natural, open-ended conversation. But can they sustain a coherent conversation for tens of minutes? Because they generate a token every frame, even while listening, they exhaust their context windows within minutes. Yet existing benchmarks evaluate only single turns or a few conversational rounds, too short to expose the long-horizon problem. We introduce the Full-Duplex Long-Horizon Benchmark (FD-LHB), which tests five recent full-duplex models on long replayed conversations and in live conversation with an LLM partner. Every evaluated model eventually either degrades or exhausts its context. Models that keep the full history often degrade before their context runs out, while sliding-window models still fall into repetition, silence, and gibberish. For sliding-window models, we identify two contributing mechanisms. Gibberish and silence are linked to Initialization-Anchor Eviction, where the sliding window discards the first KV entry. Repetition is also reinforced by the model's own recent outputs. Guided by this diagnosis, we apply two simple fixes: Initialization-Anchor Pinning (IAP) keeps that first entry, and Duplex Repetition Penalty (DRP) discourages echoing while guarding against prolonged silence. Together, they support the diagnosis and, without retraining, substantially reduce conversational collapse in replays of up to an hour while running in real time.

★ Can LLM Annotators Match Human Consistency? Benchmarking Human Agreement and Optimizing Multi-Agent with Contrastive Knowledge

Xiaopeng Ke, Zhixin Wang, Xuebo Liu, Xinyu Ma, Xing Zheng, Bowen Song — Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, New model 36%

LLM-based annotation has advanced rapidly, yet existing benchmarks remain misaligned with real-world practice. Actual annotation often involves lengthy, task-specific guidelines, and different annotators may produce divergent but reasonable judgments, especially on boundary cases, where single-labeled dataset fail to reflect this disagreement. This gap also limits existing single-agent and workflow-based annotation systems, which often lack adaptability to complex, evolving guidelines. To bridge this gap, we introduce ANT-A Benchmark, a real-world multi-annotator benchmark spanning 7 datasets across domains and modalities, with task-specific guidelines and 5 independent human labels per instance. ANT-A enables a distribution-aware evaluation, using Fleiss kappa to assess how closely model annotations align with human labeling distributions. We further propose ANT-OR, a multi-agent annotation framework that refines prompts by distilling contrastive decision rules from positive and negative feedback. Experiments show that ANT-OR improves agreement@>=2 accuracy by 4.5 points over the strongest baseline and achieves the highest alignment with human annotator distributions as measured by Fleiss kappa within 2 points. Code and benchmark are available at https://github.com/Krueske/ANT-OR.

★ Async-RBench: Benchmarking Asynchronous Task Execution in LLM Agents

Detao Zeng, Guilin Ou, Nuo Chen, Yunhao Feng, Hongliang Wang, Qinli Yang — Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 41%

Large language model (LLM) agents increasingly execute tasks asynchronously: a main agent continues working while its subagents run, and the task state may change before their results arrive. The main agent must decide which work to update or preserve and which checks to repeat, while avoiding prohibited changes. Final task success alone does not reveal whether these results were handled correctly, and natural asynchronous arrivals are hard to reproduce. We introduce Async-RBench, a benchmark for asynchronous task execution in LLM agents, with 200 executable tasks covering eight scenarios. A scenario controller regulates result delivery using predefined task-state conditions, while an Event Response Contract (ERC) specifies what each response must change, preserve, avoid, and re-verify. The Dynamic Replanning Score (DRS) weights the response process and outcome equally and is reported separately from task success. We evaluate nine agents and find DRS scores of 19.2 to 45.1, with the lowest scores on duplicate reports and subagent failures. Under our execution protocol, all nine agents obtain lower task success with asynchronous than with batched delivery (26.0% vs. 29.7% on average). For three agents, prompting the main agent to rebuild its plan after every result improves DRS by only 2.6 to 3.5 points. Together, these findings highlight the challenges of reliable asynchronous task execution and the need to evaluate how agents handle incoming results alongside end-to-end task success.

★ SCI-AGENT-EXAM: Benchmarking Agents on Scientific Research Tasks

Tianyun Zhong, Wei Wang, Jialin Li, Jianhong Tu, Liyao Li, Chuqiao Kuang — Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 69%

Large language model (LLM) agents are increasingly capable of supporting scientific workflows, but their capabilities across diverse research activities remain difficult to assess systematically. Existing benchmarks often focus on selected scientific capabilities or task types, limiting a comprehensive characterization of scientific agents' strengths and weaknesses in realistic research workflows. We introduce SCI-AGENT-EXAM, a workflow-oriented benchmark comprising 120 expert-curated tasks across 11 scientific fields and 7 research task types defined by the Scientific Workflow Task Taxonomy (SWTT). Each task requires both executable artifacts (e.g., code and scripts) and textual artifacts (e.g., reports and explanations). To evaluate the heterogeneous task outputs, we develop a hybrid evaluation protocol combining deterministic programmatic verification for executable artifacts with a calibrated, expert-validated LLM-based judge for textual artifacts. Beyond outcome assessment, we analyze progress and solution stability along execution trajectories to characterize how agents reach their final results. Experiments with 10 frontier LLM agents reveal substantial performance gaps: GPT-6 Astra achieves the highest overall score of 80.9, yet only 26.0% of its runs exceed the 0.95 success threshold. Further analyses uncover distinct capability profiles across scientific task types and substantial variation in execution efficiency and stability. Our code and data are available at https://github.com/SKYLENAGE-AI/Sci-Agent-Exam.

★ IGG: A Benchmark for Interactive GUI Grounding under Visibility Constraints

Kyeongseon Kim, Jiyeon Son, Tae-Hyun Oh — Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 47%

GUI grounding benchmarks ask whether an agent can localize a target that is already visible in a static screenshot. Real interfaces often violate this assumption: a target may be off-screen, appear only after a state change or hover, remain occluded, or become actionable after a delay. This leaves the ability to recover target visibility unevaluated. To measure this reveal step, we introduce Interactive GUI Grounding (IGG), a benchmark in which an agent must first make the target visible through interaction and then click it. IGG contains 601 tasks across 16 controllable GUI environments and seven visibility-constraint sub-types, with a compact action space of scrolling, moving, dragging, waiting, and clicking. In our human study, participants achieve 94.6% accuracy, while GPT-5.5 reaches 79.4% and the strongest evaluated open-source VLM reaches only 31.6%. To identify the source of this gap, we conduct two controlled interventions. Telling Qwen3-VL which interaction to perform raises its accuracy only from 30.8% to 33.9%, whereas performing the interaction for the model raises it to 81.9%. The bottleneck is executing the reveal interaction, not recognizing the target once it is exposed. We further show that this capability is learnable, demonstrating that visibility recovery can be strengthened consistently across model families.

★ PHYRD: A Living Benchmark for Reproducing and Extending Theoretical Physics Research

Qituan Zhang, Guo Chen, Linfeng Zhang, Wenlong Zhang — Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 41%

Large language models and agents have shown significant power in advanced physical research. In this paper, We introduce PHYRD, a living benchmark for theoretical-physics reasoning with 63 Reproduce and 43 Discover problems. Reproduce tests derivation of published results; Discover tests consequences of modified theoretical conditions. FormulaForest constructs the tasks with reference solutions and subquestion-level rubrics. Across five models, whole-problem accuracy ranges from 49.21% to 68.25% on Reproduce and from 18.60% to 74.42% on closed-book Discover. With tools, four models complete more Discover problems, but conceptual errors and missing arguments persist. Besides, our experimental results demonstrate the value of FormulaForest as synthetic data. On FrontierScience, a benchmark developed by OpenAI, it raises the accuracy of a Qwen-35B-A3B model from 61% to 72%.

★ LGMT: A Tool for Testing the Logical Reasoning Reliability of LLMs

Zenghui Zhou, Xiaoke Fang, Zheng Zheng — Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 34%

Large Language Models (LLMs) increasingly serve as reasoning components in modern software systems, yet a model that answers a logical query correctly can contradict itself when the same problem is restated in an equivalent form. Static benchmarks evaluate queries in isolation and therefore fail to expose such inconsistencies. To address this limitation, we present LGMT, an open-source tool that detects reasoning inconsistencies through metamorphic relations grounded in first-order logic. Given a source test case in premise-conclusion form, LGMT generates logically equivalent follow-up cases and checks whether the model's entailment predictions agree across the pair, without ground-truth labels. The tool provides 20 built-in metamorphic relations, supports custom ones verified by the Z3 theorem prover, and runs from a CLI or a web studio, including an offline mock mode requiring no API key. Across six widely used LLMs on a bundled benchmark of 3,091 metamorphic groups, LGMT reveals inconsistency rates ranging from 20.84% to 29.44%. LGMT is publicly available on GitHub at https://github.com/MaxwelsDonc/lgmt-tool, with a demonstration video at https://youtu.be/gou4opptu1U.

★ EvalResearchBench: Can AI Agents Build Their Own Feedback for Recursive Self-Improvement?

Yaolun Zhang, Tianyi Xu, Yujie Zhao, Jishen Zhao, Qingyun Wu, Huazheng Wang — Mon, 26 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 60%

Recursive self-improvement (RSI) relies on evaluation feedback to assess progress and guide further research, yet repeatedly running complex benchmarks is costly and slows iteration. Human experts reduce this cost by selecting benchmark subsets or designing compact suites. We ask whether AI agents can automate this design process and introduce EvalResearchBench (ERB), a benchmark for autonomous evaluation research. Given target materials, development references, candidate APIs, and fixed time and API budgets, an agent called the researcher selects or synthesizes tasks, implements graders, and revises them in pilot tests before freezing an executable evaluator for coding, co-work, and reasoning. We study 9 researchers and 13 candidate models and compare each frozen evaluator with 14 target benchmarks on score concordance and pairwise agreement. The best evaluators order about 75\% of candidate pairs as the targets do, below the 91\% ceiling set by disagreements among the targets. No researcher leads on every metric, and the best evaluator on development targets is not the best on sealed targets hidden from the researcher. A human-designed sample of public tasks remains a strong baseline, and the evaluator with the lowest recorded execution cost attains the highest pairwise agreement. Agents repair tasks and graders through pilot feedback, yet their evaluators can still truncate answers, exhaust the evaluation budget, or let a few questions dominate a domain score.

★ LAMB: A Modern Contextual Language Model for Latin

Aidan McCarthy — Mon, 26 Oct 2026 00:00:00 +0000

97% · New model 97%, Evaluation 97%

This article introduces Latin Modern BERT (LAMB), an encoder for the Latin language based on the ModernBERT architecture, pretrained in two phases on more than 74B Latin tokens using a novel morphologically informed masking regimen. We also present two corpora for the Latin language. Acervus contains 24B BERT WordPiece tokens, including over 760K samples from heavily damaged OCR transcriptions, repaired by a ByT5 model trained for Latin text repair. The second dataset, Corpus Trinum, contains a mixture of over 100M strong and weak Latin language semantic triplets. Corpus Trinum is the first Latin semantic triplet dataset, opening the door for Latin language models to be adapted to general sentence semantic similarity tasks. We construct an experimental grid using five tasks and seventeen subtasks, applying a variety of finetuning configurations and subword aggregation techniques for 15 pretrained encoders, 6 LAMB training checkpoints, and an ablation model designed to quantify the impact of different masking regimens. Using this benchmark, we rank Latin encoders and find that LAMB outperforms all other evaluated models on 16 of 17 subtasks, coming within 0.8 percentage points of state of the art on the seventeenth. We provide additional analyses relating Latin encoder performance to tokenizer behavior, syntactic structure, finetuning configuration, pretraining state, subword aggregation, and variation in source texts and domains. In these, we identify unique acquisition dynamics for Latin open class words during terminal decay phases, and establish that for some task categories, finetuning configuration selection can have a greater effect on downstream scores than additional pretraining tokens.

★ Kolibri: A Sovereign European Model on the Pareto Frontier

Mon, 26 Oct 2026 00:00:00 +0000

97% · New model 97%, Efficiency 69%

This report introduces Kolibri, an English-German mixture-of-experts transformer with 78.1B total parameters and 3.46B active parameters per token, released as open weights under the Apache 2.0 licence. We built Kolibri for sovereign and specialised deployment, with a particular focus on German and on regulated domains such as public administration, industry, and aerospace. To support these settings, we control the full model-development process, including data, architecture, training infrastructure, post-training, and evaluation. We train Kolibri on 24T tokens across pre-training, mid-training, and long-context extension, with German accounting for more than 20% of the data mix by bytes. To reach this share, we complement German datasets with more than 2T German tokens that we curate from the web with our own pipeline and generate synthetically. Post-training combines supervised fine-tuning on curated German and English data with reinforcement learning on more than 1.2M internally curated tasks, which cover reasoning, tool use, instruction following, code, and retrieval. The model activates 4.4% of its parameters per token, and among the open models evaluated in this report, it lies on the Pareto frontier of quality and serving cost in English and German. Kolibri builds on Kolibri Origin, an earlier iteration that validated our full pipeline at scale. Both models come from our Model Factory, which connects data, experimentation, training, evaluation, and deployment through version-controlled workflows.

★ Reproducible Evaluation as the Substrate for Safe Prompt Compression in Production Agents

Prabal Vashisht, Zheng Hu, Ying Yang — Mon, 26 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Efficiency 59%, Agents 46%

Evaluating conversational agents that act on live, mutable enterprise data is hard: replayed sessions drift as backend state changes, and a benchmark built today can silently go stale as the data it describes moves on. We present an offline evaluation harness for a production conversational agent that stays reproducible despite the live data it operates on changing, and that can be regenerated on demand once it goes stale. We then use this harness as the fitness function for GEPA-based prompt optimization under a correctness gate that guards against reward hacking. Applied to a production merchant-facing agent, correctness-gated compression shrinks the agent's full prompt surface by 3.5-7.6% without regressing held-out task correctness, consistently across three independent subject-model families, and for one, correctness measurably improves.

★ MedCaseAgent: From Multimodal Clinical Evidence to Rare Disease Case Reports

Mon, 26 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 95%, New model 87%

Although Large Language Models (LLMs) have demonstrated potential in clinical documentation, generating the rigorous, evidence-based reasoning required for unprecedented, atypical, and rare disease case reports remains a significant challenge. We present MedCaseAgent, an end-to-end agentic reasoning paradigm that bridges this gap by synthesizing publication-quality reports from unstructured clinical notes. MedCaseAgent leverages a hierarchical structure, integrating patient-level data mining with iterative, tool-integrated reasoning. By dynamically querying medical literature and knowledge bases, the agent maintains high clinical fidelity and assesses case novelty. The framework is inherently modular, supporting various backbone LLMs in a plug-and-play fashion. To facilitate this research, we release MedCase-150K, a large-scale multimodal clinical case-report corpus, and MedCase-Bench, a robust temporally held-out evaluation suite. We also introduce an open-source model post-trained from Qwen3.5-9B and optimized specifically for this task. Under both LLM-judged and rule-based evaluation, MedCaseAgent outperforms existing general-purpose and specialized medical LLMs and agents. Blinded expert reviews and ablations further support the system's clinical relevance and evidence grounding.

★ Every Edge Needs a Quote: Building a Quote-Grounded Knowledge Graph of Healthcare Software Adoption with Language-Model Agents

Juan Felipe Duran, Arthur Tobler, Martin Valli\`eres — Mon, 26 Oct 2026 00:00:00 +0000

97% · Agents 97%, Hallucination 91%, Evaluation 67%

Hospitals regularly buy software that clinicians and patients depend on. Because these purchases are costly, hard to reverse and can affect patient care, buyers look for evidence of which peer institutions already use a product and which systems it connects to. This evidence is public but scattered across vendor websites, press releases and third-party pages. Language-model agents can now collect it at low cost, but the correctness of the extracted claims is difficult to establish, and a wrongly listed customer can mislead buyers and misrepresent an organisation. We present a pipeline that builds a knowledge graph of customer and integration relations from the open web under one rule: every edge must be supported by a verbatim quote from its source page. Pages come from a deterministic crawl of each vendor's site and a budgeted ReAct agent that searches the rest of the web. A permissive extractor proposes relations with quotes, and a layered grounding stage decides which to keep, using a referent step that reads the whole page when a quote names only one side, deterministic checks, and a panel of five model calls voting on nine criteria. On 61 vendors, the pipeline reaches 0.94 precision and 0.71 recall against a page-level key covering the 6,428 candidate relations proposed by its extractor, whose verdicts were drafted by one model and re-checked with a second. Against the golden reference, the marketplace's own curated list, it scores lower, but 89% of the relations it keeps that the reference lacks are stated on a page. A confidence score built from the panel's votes keeps more correct relations than the fixed rule at the same precision, and reaches 0.90 precision at 0.90 recall. We also discuss how narrow, bounded model calls, string-matched evidence and voting address long-context degradation and run-to-run variance.

★ Efficient Omni-modal Large Language Model Inference via Per-Query Modality Selection

Shuqi Liu, Yuzhou Cao, Luke Ong — Mon, 26 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, Evaluation 56%, New model 45%

Omni-modal large language models (Omni-LLMs) can jointly process a broad range of modalities, such as text, audio, and video, enabling stronger multimodal understanding and reasoning capabilities, but often at substantial inference cost. We observe that the benefit of additional modalities varies across queries, indicating that a fixed modality configuration can be inefficient. Motivated by this, we study query-wise modality selection, which keeps the underlying Omni-LLM fixed while adaptively selecting which modalities to provide for each query. We formulate the problem as an accuracy-cost optimization problem and develop a cost-sensitive learning approach for modality selection. We evaluate our approach on WorldSense using Qwen2.5-Omni and Qwen3-Omni under three inference-cost measures: input tokens, latency, and monetary cost. Across both models, query-wise modality selection consistently improves the accuracy-cost trade-off and can even outperform the standard setting in which all available modalities are provided to the Omni-LLM. For Qwen2.5-Omni, our cost-sensitive approach matches or exceeds this full-modality accuracy while requiring only approximately 41% of the input tokens, 39% of the latency, and 61% of the monetary cost.

★ SPECMEM: Accelerating Agentic Tool Calling via Live Memory Management

Zihe Song, Wenting Zhao, Ye Liu, Yuanjie Zhu, Ke Xu, Shanghao Li — Mon, 26 Oct 2026 00:00:00 +0000

96% · Agents 96%, Efficiency 95%, Evaluation 50%

Tool-calling agents are mostly optimized for accuracy; in deployment, however, their growing cost is latency: every tool call is decoded autoregressively. Speculative decoding accelerates this without changing the output, and retrieved past calls make strong drafts under real use, where each user's behavior tends to repeat. Yet existing systems keep no cross-session memory, or build the draft store once and never update it. We propose SpecMem, a tool-call memory kept alive and managed throughout deployment: verified calls are written back online into bounded per-user stores with LRU eviction, and a side-effect gate defers irreversible calls until verification confirms. On multi-session, multi-user query streams served by three open-weight models, a frozen store steadily loses acceptance while SpecMem holds it: a 27% gap in mean accepted tokens (+11% to +34% across benchmarks, tracking how much novel work arrives after the store is built) and up to 1.74x wall-clock speedup. Ablations attribute the gain almost entirely to keeping the store live; local per-user scoping adds no acceptance but closes a measured cross-user leakage channel, and gating on declared side effects, rather than on retrieval confidence, is what eliminates irreversible mistakes.

★ Who Verifies the Tools? How Warped Formatters Mislead Agents and Their Verifiers

William Jeffries — Mon, 26 Oct 2026 00:00:00 +0000

96% · Agents 96%, Hallucination 86%, Evaluation 66%

Agents and verifiers often rely on the text a tool prints, and that text can misrepresent what the tool actually did, even when the tool's computation is correct. We saw this happen in a production RAG research system used by the Pilsudski Institute in New York, where tool output formatters misled an agent into giving wrong or incomplete information to users. Each tool did its job correctly, but its output omitted search limits, result caps, or identifier namespaces. We also recreated an unlabeled estimate caught in code review before release. We studied whether these formatters misled agents and verifiers and how much the repaired wording helps. We fed tool results to agents using two formatter versions: one with the key omissions and one with all missing details clearly stated. That difference alone changed what the agents claimed and what the verifiers accepted. Unsupported agent claims dropped substantially for three of the four failure types. In one test, verifiers went from accepting incorrect answers 90 out of 90 times to rejecting them 90 out of 90 times. A control showed the verifiers could catch the errors once they were given the missing facts and asked to judge correctness. We include the audit prompt we used with Codex and Claude Code in an exploratory demonstration, written with prior knowledge of our own system's defects. When pointed at an old version of our own system, it rediscovered the defects that were still live and surfaced twelve more we had not catalogued. When pointed at OpenClaw and Hermes, two of the best-known open-source agentic systems, it found warped formatters in both. We opened PRs to fix both, and the OpenClaw fix has already been merged. Practitioners can run the same prompt on their own systems today.

★ SOPO: Socratic Guided Policy Optimization for Span-Level Hallucination Detection

Jiawei Dong, Zilong Bai, Pengtian zhu, Peng Zhou — Mon, 26 Oct 2026 00:00:00 +0000

96% · New model 96%, Hallucination 95%, Evaluation 47%

Large Language Models (LLMs) have gained widespread adoption in various natural language processing tasks, but suffer from hallucination issues where they generate unfaithful or inconsistent content. While Reinforcement Learning with Verifiable Rewards (RLVR) has shown promising applications in reasoning tasks, its application to span-level hallucination detection is challenged by localization ambiguity, reward sparsity on hard examples, and uncontrolled length inflation. To address these issues, we introduce Socratic Guided Policy Optimization (SOPO), a multi-turn RL framework for robust hallucination detection. The key innovation of SOPO is a Socratic feedback mechanism, where a teacher model guides the reasoning trajectories during rollout by providing indirect hints inferred from the ground truth. This teacher-guided approach enables the model to generate more effective reasoning trajectories for hard examples, thus alleviating reward sparsity. Furthermore, we incorporate Turn-based Advantage Scaling to penalize verbosity bias, encouraging the model to develop efficient and autonomous reasoning capabilities. Empirical results on RAGTruth demonstrate that SOPO (4B) achieves state-of-the-art performance, surpassing leading proprietary models (e.g., GPT-5) and standard RL baselines, while obtaining higher reported mean scores than GRPO on four out-of-domain benchmarks, with mixed evidence against CAPO. We release our training and evaluation code alongside the 1.7B and 4B SOPO models at https://github.com/patsnap/SOPO.

★ RxnExBench: Benchmarking Vision-Language Models for Structured Chemical Reaction Extraction from Scientific Literature

Jiaxing Sun, Junyuan Gao, Ruikang Zhang, Hao Zheng, Lingli Ge, Xiao An — Mon, 26 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

Screen Tables and Substrate Scopes are information-dense figures widely used in synthetic chemistry literature, yet automated extraction of structured reaction data from them remains challenging. We introduce RxnExBench, a multimodal benchmark for fine-grained reaction-level evaluation. Built from high-impact chemistry journals, it contains 3,565 expert-annotated reactions and 2,361 molecular structures. The dataset adopts a unified schema with 15 reaction roles and 26 core fields, supporting downstream AI-for-Chemistry tasks such as retrosynthetic planning, yield prediction, and reaction condition recommendation. RxnExBench further provides a reasoning-driven difficulty taxonomy, a modular annotation pipeline, and the Oracle-MolRec protocol, which disentangles molecular recognition from higher-level reasoning. Evaluation of 9 state-of-the-art VLMs shows that performance degrades substantially as schema complexity and reasoning demands increase. Under the Oracle-MolRec setting, results further reveal that accurate molecular recognition alone is insufficient, with multimodal integration and chemistry-aware reasoning remaining the primary bottlenecks. We release all data and code to facilitate future research in chemistry-aware multimodal understanding and AI-for-Chemistry dataset construction. The dataset and code will be released at https://github.com/opendatalab/RxnExBench.

★ Do Language Models Trust Their Own Justifications? A Study on Functional Consistency

Alisson Rosa Pereira, Bruno Iochins Grisci — Mon, 26 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

Large Language Models (LLMs) have been widely adopted in text classification tasks, where they not only output class predictions but also generate explanations that highlight the tokens deemed most relevant for reaching the predicted label. Yet it remains unclear whether these highlighted elements are behaviorally consistent with the model's own predictions under targeted interventions, which we refer to as functional auto-consistency. While much of the literature evaluates the textual plausibility of such explanations, few studies assess their functional consistency with the model's actual behavior. In this work, we propose an experimental framework based on the principle of auto-consistency: if a model identifies certain tokens as decisive, then isolating, removing, or semantically inverting them should produce systematic and interpretable changes in its predictions. We operationalize this evaluation through sufficiency, comprehensiveness, and counterfactuality metrics, and conduct experiments on IMDB and Steam reviews across both closed-source (GPT-4o-mini) and open-source LLMs (Gemma3, Granite8B, DeepSeek). Results show that GPT-4o-mini follows the expected progression across all metrics, Gemma3 and Granite8B maintain coherence under sufficiency but lose consistency under more demanding interventions, while DeepSeek variants display structural deviations, either failing to preserve sufficiency or overreacting under comprehensiveness and counterfactuality. These findings show that explanation reliability varies across LLM families and scales, with smaller models displaying contradictions and larger ones exhibiting over-sensitivity. By combining sufficiency, comprehensiveness, and counterfactuality, our approach provides a systematic methodology for assessing the functional consistency of LLM self-explanations.

★ HarnessMM: Recursive Self-Improvement for Harness Token Efficiency

Yuhua Jiang, Chang Xu, Feifei Gao, Li Zhao — Mon, 26 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, Agents 86%, Evaluation 55%

Agent harnesses repeatedly invoke reasoning after tool feedback, even when an earlier call has already fixed the next action. We introduce HarnessMM (Harness Minus Minus), a method for improving harness token efficiency through guarded decision reuse and recursive self-improvement (RSI). RSI iteratively refines companion harness configurations from training feedback, with validation selecting the configuration for deployment. A controller at the model-tool boundary reuses resolved decisions to replace later reasoning callbacks while actor weights remain fixed. Executable Decision Graphs store exact actions, literal arguments, causal dependencies, and public guards. Guarded Decision Reuse activates these persistent decisions from observed receipts; Certified Callback Substitution executes them through the native interface and records removal of the registered slow callback. New semantic decisions, unsupported evidence, and selector abstention invoke the original actor again. The representation restricts what feedback may change, while complete-task measurements establish quality and cost. On HumanEval, HarnessMM preserves Codex success at 90.82\% while reducing Tokens/task (k) from 66.72 to 47.02 (29.54\% saving). For DeepSeek Harness, success rises from 71.43\% to 85.71\% while Tokens/task (k) fall from 96.28 to 11.62 (87.93\% saving). A separate MBPP ablation measures 26.91\% lower cost than single-use graphs and 19.52\% lower cost than retained callbacks. These results establish guarded decision reuse as a practical way to reduce repeated reasoning in existing harnesses.

★ Latent Coherence: Do Frontier Language Models' Probability Judgments Obey the Law of Total Probability?

Peter Potash — Mon, 26 Oct 2026 00:00:00 +0000

95% · Evaluation 95%, Hallucination 39%

LawZero's Scientist AI trains a Predictor toward the Bayesian posterior, with an objective that penalizes inconsistent probabilities among latent explanations. We ask whether current frontier language models, with no such training, are already consistent in this sense zero-shot: do their probability judgments satisfy the law of total probability, $P(X)=\sum_z P(z)P(X|z)$, when the latent $z$ must be inferred from records in the context? In separate single-turn calls, models are asked for $P(X)$, for $P(z)$ over the latent's states, and for $P(X|z)$; a causal family applies the same test to $P(y|do(x))$ and the back-door adjustment. Every query has an exact ground-truth probability, and a subset of items is asked in three wordings to separate persistent incoherence from rewording noise. Across seven models, GPT-6.1 Sol and Claude Opus 5.5 are the most coherent (mean absolute residual 0.08 and 0.12 log-odds, about 2 percentage points). Claude Haiku 5.5, Gemini 3.8 Flash and DeepSeek V4-Pro show persistent incoherence (0.27-0.35 log-odds) that is mostly item-specific rather than a shared offset. Within every model, larger residuals go with larger errors against the truth (Spearman 0.40-0.62), so the residual, which needs no ground truth, flags unreliable answers. Reasoning length does not predict coherence, across models or within a model across rewordings of the same question. Simple consistency checks pass almost everywhere; the failures are specific to reasoning through a latent variable. Code and data: https://github.com/ppotash/latent-coherence

★ TACS-RF: Tool-driven Agentic Circuit Synthesis of RF and mmWave Modules

Tristan Liang, Aravind Nagulu — Mon, 26 Oct 2026 00:00:00 +0000

95% · Agents 95%

We present the framework of TACS-RF - a Tool-driven Agentic Circuit Synthesis of RF and mmWave modules. In this framework, a large-language-model (LLM) coding agent (Claude Opus 4.8) directly drives industry-standard electronic-design-automation (EDA) tools such as Cadence Virtuoso/Spectre, Calibre, Cadence EMX, and Ansys HFSS to design RF and mmWave modules. We validate the framework through the agentic design of a 240-251 GHz sub-THz transmitter in a 90 nm SiGe BiCMOS process. The agent semi-autonomously designed the sub-blocks including a 60 GHz push-push Colpitts VCO with second-harmonic extraction, multistage differential cascode power drivers, a push-push frequency doubler, a differential bowtie antenna, and all the interstage matching networks. The transmitter chain with parasitic-extracted devices and EM-verified passives delivers +5.1 to +6.3 dBm across an 11 GHz varactor-only tuning range, phase noise of -80 to -88 dBc/Hz at 1 MHz offset while consuming a DC power of 361 mW. We describe the agent-driven design flow and the resulting division of labor between human and agent, taking a step toward a "text-to-ASIC" future for designing RF/mmWave hardware.

★ PRISM: Promoting Sampling Diversity of Large Language Models with Implicit Stochastic Modulation

Qixin Xu, Zihan Zhu, Xiaojiang Liu, Bohan Lyu — Mon, 26 Oct 2026 00:00:00 +0000

95% · New model 95%, Efficiency 49%, Evaluation 36%

Large language models (LLMs) exhibit mode collapse, where their output distribution concentrates on familiar semantic modes. When diversity disappears, repeated sampling becomes semantic repetition rather than discovery, making rare but stronger solutions unlikely to surface and leaving self-evolution and continual learning difficult to sustain. Most remedies intervene only at inference time, shaping sam- ples without making diversity a model capability. We propose PRISM, a training-time framework that internalizes diversity within the model. PRISM combines (i) submodular curation that pairs each prompt with diverse-yet-correct multi-teacher solutions, and (ii) stochastic embedding modulation, which turns special-token perturbations into implicit signals for sampling distinct outputs. A theoretical analysis establishes the stability of this mechanism. On WritingPrompts and BookMIA, PRISM moves the quality-diversity frontier outward, achieving the best generation quality in all model-benchmark settings while maintaining the strongest or near-strongest semantic diversity across two open-weight model families. It also improves sampling efficiency on math and code benchmarks. Beyond inference, PRISM yields a stronger foundation for reinforcement learning, sustaining diverse rollouts and mitigating the diversity collapse commonly observed during RL training.

★ Shorter Is Not Cheaper: A Controlled Comparison of Declarative Animation Specifications and Free-Form Code for LLM-Generated Explanations

Jiantang Huang — Mon, 26 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

Chat assistants increasingly answer conceptual questions with interactive, animated explanations, generated either as free-form HTML/JavaScript or as declarative specifications that a fixed runtime plays. The declarative route is widely expected to be cheaper, faster and more reliable, because its outputs are shorter and constrained. We test this expectation in controlled comparisons whose hypotheses and analysis plans were frozen before the runs. We designed ExplainSpec, a time-indexed declarative animation format, and an observation contract that makes every generated explanation programmatically inspectable; validated, pacing-agnostic checkers score 20 explanation requests from five domains. Three Claude models (360 jobs) and, in a follow-up, the open-weight Qwen3-14B with thinking enabled and disabled (240 jobs) answered every request in both formats. For every model the specifications were 25-48% shorter in answer tokens (output excluding reasoning), yet with reasoning enabled total output fell significantly only for the largest Claude model and no model generated significantly faster. The reasons differed. The Claude models reasoned far more before writing specifications (exploratory; for the two larger models, from 15-20% to 85-88% of generations). Qwen3-14B reasoned about as much in both formats (a pre-registered test of this shift was not supported), and its shorter answers saved little against a reasoning trace more than twice as long. Nor were specifications more reliable: free-form pages were valid after one repair round in 95-100% of jobs for every model, whereas Qwen3-14B produced valid specifications in only 33% of jobs and passed all semantic tests in 17% (HTML/JavaScript: 62%). For reasoning models, a shorter format alone therefore does not make generation cheaper, faster or more reliable, and a format learned in context can cost reliability.

★ Correct Answers, Closed Gates: SolveBind for GUI Agents at the Verification Barrier

Xinhao Song, Haoxuan Peng, Su Su, Hongliang Wu, Sirui Song, Gongshen Liu — Mon, 26 Oct 2026 00:00:00 +0000

95% · Agents 95%, Evaluation 54%, Efficiency 30%

Websites and applications deploy CAPTCHAs and anti-automation mechanisms that use behavioral signals such as interaction trajectories and timing, not outcomes alone, to distinguish humans from automated agents. We call these protected GUI environments, where correctly solving a verification task is insufficient for end-to-end task completion. Agents must turn solver outputs into accepted interactions and adapt execution to runtime feedback under time constraints. We introduce SolveBind, a GUI-agent framework combining hierarchical solving with runtime supervision. A hierarchical solving tree decomposes verification into interdependent Perceive, Ground, Solve, Act, and Verify capabilities. A runtime supervisor grounds action commitments in tool-derived evidence, revises them as evidence arrives, and mediates execution under explicit recovery budgets. Across nine CAPTCHA task families and nine backbone models, SolveBind achieves the highest success rate on eight backbones, outperforming the strongest baseline by 13.3-34.5 percentage points. It achieves the highest success rate on all six AndroidWorld backbones. In a paired GPT-5.5 analysis, MobileAgent produces correct answers but fails dynamic verification on 13 of 90 tasks; SolveBind succeeds on seven of these tasks. With automatic submission, removing binding reduces success from 82.22% to 71.56% and increases latency and token usage. Recovery caps reduce execution cost with a small observed success trade-off: removing them raises success from 82.22% to 83.56%, while increasing latency and token usage to 3.7x and 2.4x their full-supervision values, respectively. These results support evidence-grounded execution control while identifying a success-cost trade-off in recovery and failure boundaries of heuristic submission rules.

★ Solvent: Long-Horizon Reasoning for Multi-Step Retrosynthesis

Yifei Wang, Botian Wang, YiFei Wang, Jiangtao Feng, Xinbo Zhang, Yawen Ouyang — Mon, 26 Oct 2026 00:00:00 +0000

95% · New model 95%, Agents 70%, Evaluation 39%

Multi-step retrosynthesis requires recursively transforming a target molecule through successive reaction steps into purchasable building blocks, and recent work has begun to draw on large language model (LLM) reasoning to tackle the task. In its latest evolution, LLM reasoning has entered the long-horizon regime: models pursue objectives across interdependent steps, interact with environments, and adapt to feedback. This shift has markedly advanced LLM capabilities, as illustrated by breakthroughs in AI-driven mathematical research. At this frontier, we introduce Solvent, an 8B LLM that successfully extends long-horizon reasoning to multi-step retrosynthesis. Through a two-round post-training curriculum, Solvent learns to interleave iterative synthesis planning, external availability verification, and adaptive route revision. Experiments show that, following a $2.1\times$ performance boost from RL, Solvent scores 61.86% across four benchmarks, surpassing mid-scale language models and prior search-based retrosynthesis systems, and rivaling frontier trillion-parameter LLMs.

★ Beyond Gain: Measuring Skill Internalization for Agent Skill Retirement

Yanze Wang, Pengfei Yao, Tianyi Sun, Hongda Chen, Yunyun Han, Jun Sun — Mon, 26 Oct 2026 00:00:00 +0000

95% · Evaluation 95%, Agents 40%

LLM agents retrieve skills from public hubs that hold up to about 100,000 entries and serve many models. As models improve, a hub must find the skills a given model has internalized, meaning it can do without them, and retire them for that model, yet internalization has not been measured with known reliability. Published rules that compare runs with and without a skill retire on the gain alone. The gain is near zero both for an internalized skill and when neither the model nor the skill delivers the capability, a state we call uncovered. We instead withhold the skill, score the model on a task synthesized from it, and trust a high unaided score only when another model gains from the skill, since nearly half of high scores come from tasks no model benefits from. Under four models, a published gain rule applied with one task per skill removes over half of the uncovered skills. On SkillsBench, 28 of its 31 retirements rest on probes no run passed in full, while retiring on the unaided score costs no detectable reward. Unaided verdicts reproduce across tasks more often than the gain rule's (Cohen's kappa 0.42 against 0.23) but differ across models, so a hub must read them per model. We release 1,251 skills with tasks and item scores under four models.

★ Beyond Final Outcomes: Quantifying Claim-Level Information Dynamics in Multi-Agent LLM Interaction

Haocheng Jiang, Herun Wan, Yongyuan Peng, Minnan Luo — Mon, 26 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

When LLM-powered agents interact, information is repeatedly shared, expanded, and forgotten, yet multi-agent evaluation typically focuses on final task outcomes. We study such interaction as claim-level information dynamics and propose RIVER, which decomposes agent outputs into atomic claims and evaluates three complementary dimensions: convergence, exploration, and retention. Convergence captures cross-agent claim correspondence and stance alignment; exploration captures semantic breadth and extension beyond the source; and retention tracks source preservation and the persistence of new claims across stages and lineages. We evaluate five interaction strategies on 40 controlled news-generation cases. Repeatedly pooled interaction promotes convergence but narrows the information space, whereas distributed relay yields broader, less shared information. Further comparisons show that method-specific constraints increase exploration while weakening convergence and retention. Our findings suggest that collective information dynamics are better explained by underlying information-flow mechanisms than by simple downstream performance.

★ LeafSkill: Efficient Leaf-Level Action Grounding for Recursive LLM Agents

Shen Liu, Ze Zhou, Zhenguo Xu, Shaopu Wang, Yike Gao, Chunlei Wang — Mon, 26 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, Agents 70%, Evaluation 44%

Recursive large language model (LLM) agents solve long-horizon tasks through decomposition, execution, and environmental feedback. Yet after producing an atomic subtask, they often invoke the same large model again to translate a familiar instruction into an executable action. We observe that high-level planning is task-specific, whereas leaf-level action grounding recurs across tasks and can be checked against environment-admitted actions. We introduce LeafSkill, an external reuse layer that retrieves semantically related grounding experience, binds action templates to current objects and admissible actions, and uses a lightweight 1.7B selector to choose a bound action or defer the unchanged subtask to the original agent. Skills remain outside the model parameters, allowing a frozen selector to use post-training additions. Under lexical and syntactic paraphrases, exact matching retrieves none of the reference actions, while semantic retrieval preserves 96.0% and 97.0% candidate coverage; the frozen selector adopts 97.34% of unseen skill types. On an independent confirmation set of 200 Robotouille tasks, LeafSkill completes 137 tasks versus 121 for ReCAP (68.5% vs. 60.5%, exact McNemar p=.0259). On the 106 tasks completed by both, it reduces Qwen3.5-27B tokens by 14.42% and calls by 14.08%. A separate held-out repeat yields a consistent 9.0-point success gain and 18.98% token reduction. Reusable leaf grounding can thus improve task completion while reducing host-model inference without replacing high-level planning.

★ When Agents Fail: Model or Harness--Who's to Blame?

Shao Zhang, Hangfan Zhang, Yuanyuan Shi, Hao Li, Min-Le Su, Kangcong Li — Mon, 26 Oct 2026 00:00:00 +0000

95% · Evaluation 95%, Agents 82%

LLM agents combine models with execution harnesses, so agentic benchmark performance reflects their joint behavior. We argue that harnesses should therefore be evaluated through their interactions with underlying models. We develop a framework combining five complementary measures: (i) Valid Completion Rate (VCR), how reliably agents finish normally and deliver evaluable results, irrespective of task correctness; (ii) Model-Harness Fit, how well a harness supports task achievement through valid completion for a given model; (iii) Cross-Model Fit, a harness's overall suitability across evaluated models; (iv) Interaction Residuals, pairing-specific execution strengths and weaknesses hidden by component-level summaries; and (v) Token Load, the token demands of agent execution, including failed attempts. We evaluate four general-purpose harnesses (OpenCode, Codex CLI, Pi CLI, and DeepAgents) with four core models, plus two fixed native model-harness systems, on six benchmarks covering context management, tool use, long-context planning, multimodal interaction, and artifact handoff. Including native references and stress conditions, 111 evaluation settings yield 52,168 trajectories. Harness rankings vary across models and benchmarks; task performance and valid completion can favor different pairings, and similar task scores can accompany distinct execution profiles. OpenCode ranks first in cross-model performance, while DeepSeek V4 Flash Preview maintains consistently high Mean VCR across harnesses on the shared four-benchmark panel. These findings support evaluating model-harness interactions, training transferable harness-use capabilities in models, and harness designs adapted to the model and task.

★ Accordion: Elastic-Depth Inference via Fixed-Prefix Layer Looping

Yudong Zhang, Wei Ding, Kaili Zheng, Yuchen Ding, Zhenyu Zhang — Mon, 26 Oct 2026 00:00:00 +0000

95% · New model 95%, Efficiency 70%, Evaluation 32%

Large language models serve applications with diverse inference budgets, but training and maintaining separate models for each budget incurs substantial costs. Recent Looped Transformers enable flexible inference computation through parameter reuse, but maintaining strong performance across different inference budgets remains an open challenge. We introduce Accordion, a mixture-of-experts-based model that supports five inference budgets through fixed-prefix layer looping. Each budget follows a pre-defined execution path, while the minimum budget retains the unlooped model. Joint pretraining samples one budget per micro-batch and optimizes its next-token prediction loss without additional cross-budget consistency objectives. We further preserve these inference budgets through supervised fine-tuning and reinforcement learning, where parameter-group merging and budget-matched multi-teacher on-policy distillation integrate three domain experts into one model. After 3.6T tokens of pretraining, increasing inference depth improves Accordion-6B-A0.5B's five-category macro-average across 37 tasks by 3.01 percentage points, with the middle budget capturing 89% of this gain using half as many additional layer executions. The final model surpasses each domain expert on its corresponding task-and-budget average, while achieving strong instruction-following performance across all budgets and retaining depth-dependent gains in mathematical reasoning and code generation.

★ NgramRL: Accelerating RL Post-Training for Structured Generation via Rollout-Aware N-gram Speculative Decoding

Nilesh Prasad Pandey, Jangseon Park, Jason Kong, Onat Gungor, Tajana Rosing — Mon, 26 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, New model 78%

Reinforcement learning (RL) post-training improves the reasoning capabilities of large language models, but its cost is dominated by autoregressive rollout generation, which can exceed 70% of training time. Existing speculative-decoding methods are poorly suited to this setting: model-based drafters consume scarce GPU memory and may become misaligned as the policy evolves, while existing model-free methods do not explicitly exploit repetition across training epochs. We observe that recent rollouts of the same prompt provide a high-quality, self-updating draft source. Based on this insight, we propose NgramRL, a per-prompt n-gram drafter constructed from bounded recent rollout history. NgramRL requires no auxiliary model, drafter training, or persistent GPU-resident drafter state, while target-model verification preserves the rollout distribution. To hide CPU cache lookup and scheduling, we further introduce an adaptive fan-out prefetcher that predicts likely verification outcomes (j, b), consisting of the accepted draft length j and target-emitted continuation token b. A causal budget allocator concentrates prefetches on high-probability outcomes, while uncovered outcomes fall back to sequential lookup without changing committed tokens. Across two model families and three reasoning and structured-generation workloads, NgramRL commits 3.60-7.77 tokens per target-model forward in the warm-cache regime, up to 1.99x more than SuffixDecoding and 7.0x more than prompt-lookup decoding. On Spider, the prefetcher covers 96.6% of Qwen3-8B outcomes at a candidate budget of 16 and improves measured rollout throughput by 1.38x over synchronous NgramRL, approaching its 1.40x overlap ceiling.

★ UserArena: Benchmarking Interactive User Simulation for Long-Horizon Shopping Agents

Jujia Zhao, Jiakai Tang, Jing Yu, Wen Chen, Jian Wu, Han Zhu — Mon, 26 Oct 2026 00:00:00 +0000

95% · Evaluation 95%, New model 79%, Agents 37%

Agentic recommendation pursues shopper goals through sequences of context-dependent decisions. Developing and evaluating such policies requires faithful recommendation world models that capture both the interactive shopping environment and users' behavioral responses over long interactions. Existing resources rarely provide long-horizon, fine-grained real-user trajectories or support trajectory-level evaluation and offline validation against production A/B results. We introduce UserArena, a real-world e-commerce interactive benchmark containing 64,729 session trajectories from 11,010 users, spanning 566,102 items and 13 action types. UserArena pairs these trajectories with an interactive environment and supports offline validation against outcomes from a controlled production A/B test, enabling five-level evaluation from output validity and local decisions to sequential behavior, personalization, and outcomes. To improve simulator fidelity over these long-horizon interactions, we further propose On-Policy Trajectory Distillation (OPTD), which uses trajectory-level intent to provide dense teacher targets on student-generated rollouts. Experiments show that OPTD achieves the strongest overall fidelity among the evaluated simulators under UserArena's evaluation protocol and most closely reproduces the production policy effect in offline simulation.

Learning to Be Monitored: Causal Bounds and Intervention-Aware Post-Training Recipes for Chain-of-Thought Monitorability

Qiang Fu, Xinyue Li, Andre Wibisono, Arman Cohan — Mon, 26 Oct 2026 00:00:00 +0000

94% · Evaluation 94%, New model 77%

Chain-of-thought (CoT) monitoring aims to detect potentially harmful behaviors in large language models (LLMs) from their CoTs. Intervention-based evaluations measure this property by asking whether a monitor can detect the role of an intervention from the CoT when the targeted behavior occurs, typically using minimal-criterion sensitivity ($\mathsf{TPR_{mc}}$) proposed by~\citep{guan2025monitoring}. However, a target behavior observed under intervention is not necessarily caused by it, and we show that this metric can overestimate how often a monitor detects target behaviors actually caused by the intervention. Specifically, we establish $\mathsf{TPR_{mc}}$ as an upper bound on this causal detection rate and derive a complementary lower bound, which provides a tighter evaluation framework for CoT monitorability. With this framework, we study whether LLMs can be trained to make their CoTs easier to monitor. We first construct a training dataset of paired control--intervention tasks spanning multiple intervention types. Using the dataset, we design and apply three intervention-specific training methods based on off-policy self-distillation, reinforcement learning, and on-policy self-distillation to train separate variants of each base model. Across various model-monitor configurations, all three methods generally improve CoT monitorability over the corresponding base models under our extended evaluation framework. Our code is available at https://github.com/xinyue-lxy/learning-to-be-monitored.

Phronix: The Capability-Centric Learning Paradigm for Agents

Mon, 26 Oct 2026 00:00:00 +0000

94% · Agents 94%, Efficiency 81%, Evaluation 46%

Scientific computing increasingly relies on large language model-based agents for performance engineering tasks such as instrumentation, analysis, and workflow porting. However, agent effectiveness depends heavily on the surrounding harness, which typically relies on manually written skills and curated memory. This limits autonomous learning, forcing agents to repeatedly rediscover solutions rather than retain and reuse knowledge from prior executions. We introduce Phronix, a capability-centric learning paradigm that enables agentic harnesses to acquire, validate, and refine reusable capabilities from their own executions. Phronix operates through a predictor-validator-corrector loop: MCP tools and associated skills propose solutions, tool executions validate them, and observed outcomes refine the stored capabilities. It distills every observed interaction into a capability store and researches a failing task rather than abandoning it. First, such learned capabilities solve a recurring task on the first attempt in 93-95% of cases across every model, significantly reducing solution non-determinism and improving reliability. Second, Phronix turns an agent's own execution into 7.6-30.8 durable capabilities per task without user instruction, and creation falls below two once the needed capability exists. Third, warm-start runs consume only 0.38-0.40 million tokens per task, which is three to four times less than a single baseline attempt, while a cold-start costs no more than the baseline. On the 1000 Genome Pegasus workflow, instrumenting a production application for I/O performance analysis, Phronix passes all 11 correctness checks against six for the baseline harness, and also lifts the task solving rates of two cheaper models from 29% and 27% of SWE-bench tasks solved to 91% and 87%, respectively, matching the performance of the most expensive evaluated model.

What the Query Remembers: Query Abstraction Governs Lexical and Dense Retrieval of Classical Arabic Poetry

Omar Farouk Zouak, Selma Mani, Larbi Said Chikh, Raid Abderrezak Ouahioune — Mon, 26 Oct 2026 00:00:00 +0000

94% · Evaluation 94%

How well a poetry archive can be searched depends on what the user still remembers, not only on what the model understands. We show this for Classical Arabic poetry, benchmarking nine dense encoders and BM25 over two corpora, three indexing modes and two evaluation protocols. Which retriever wins is governed by how much of a poem's wording the query still carries. On tip-of-the-tongue queries BM25 leads every dense encoder, including an 8B state-of-the-art model (0.746 vs. 0.717 Recall@5); on abstract queries it falls to eighth out of ten (0.392 vs. 0.748). Editing the wording of the same queries shows that this ranking largely reflects query composition: adding one remembered half-verse raises BM25 by 38 points, rewriting the quoted words by hand leaves BM25 level with the 8B encoder, and BM25's sensitivity to remembered wording is significantly steeper than that of the strongest dense encoder. The same axis controls explanation augmentation, whose effect swings from +37 to -39 points. The mechanism splits by system type: lexical containment predicts BM25's behaviour, while embedding proximity predicts dense encoders. Finally, the field's two standard protocols, known-item retrieval and thematic relevance, dissociate on the same systems, a result reproduced by two blind human annotators and by two open-weights judges, one of them under a rubric with no literal-match criterion. A benchmark that reports one protocol, or one query register, can give a materially wrong picture. Code and data are available at https://github.com/Ellzo/what-the-query-remembers.

Beyond Data Schedule: Fast LLM Pretraining on Data Mixtures by Loss-Reweighting

Jiazheng Li, Huaqing Zhang, Jing Xu, Tianle Cai, Jingzhao Zhang — Mon, 26 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 41%

Large language models often acquire specialized capabilities through successive pretraining stages, but these gains can come at the cost of forgetting earlier knowledge. In this work, we investigate whether such trade-offs can be improved by changing how strongly each domain contributes to learning while keeping its data exposure fixed. We find that, in controlled knowledge--mathematics experiments, loss reweighting produces a single checkpoint that surpasses the best knowledge and reasoning scores attained along either stage trajectory. This reveals an independent control that data-mixture ratios alone do not capture. Building on this finding, we propose LARE (Loss-Adaptive Reweighting under fixed Exposure), which automatically adjusts domain weights from running loss statistics. Motivated by a theoretical analysis of gradient noise, its normalized inverse-square-root rule balances relative loss scales without changing the sampled mixture or requiring additional runs to search for weights. Across knowledge, mathematics, and code, \Method{} improves all five evaluated benchmarks for a 1.5B model trained on 240B tokens, averaging +7.84 pp over a matched direct mixture, including gains of +17.06 pp on GSM8K and +9.00 pp on MATH. These findings demonstrate that adjusting domain loss scales can improve joint capabilities without additional domain exposure.

Measure Before You Train: Typed Decision Models vs. a Classical Baseline vs. a Generative LLM for Routing a Personal Multilingual Inbox

Federico Lucca — Mon, 26 Oct 2026 00:00:00 +0000

94% · Evaluation 94%

Personal email agents decide, for every message, what it is, whether it is urgent, whether to draft a reply, create a task, or ignore it. Their error rate is rarely measured: the stored labels were written by the system itself. We present Foyer, a self-hosted personal inbox agent in daily use, and a measurement protocol (stratified sampling, frozen snapshots, thread-grouped splits, pre-registered hypotheses, human labels in the agent's review loop) that yields ground truth before any training. On 739 real Italian/English emails labelled by the owner (231 unbiased, 500 adversarially sampled), we compare nine systems under thread-grouped cross-validation: the production keyword rules, a TF-IDF + logistic-regression baseline, local typed decision models (Tev1, Nimble), a hosted decision model (Nimble v4-max), and hosted generative LLMs (Qwen2.5-7B, Gemma 4 26B/31B). Four results. The production rules score 28% on message type against a 58% majority class: the field shown most prominently carried no information. With ~200 labels, logistic regression has the highest accuracy on every field (type 0.79, urgency 0.74, needs-reply 0.83, is-noise 0.90), though it never predicts the two rarest classes and is statistically tied with Gemma 4 on type; few-shot prompting does not close the gap. The sender's relationship (client, supplier, service) is recovered from other emails of the same contact with 98.6% accuracy where the contact has been seen, and is not inferable from a single message by any model: it belongs in the contact graph, not the prompt. Finally, accuracy is not the product metric: composed into actions, the most accurate system most often ignores emails needing action (18% vs 4% for the best LLM); a cost-chosen threshold on its noise probability halves that rate and makes it the cheapest system.

CuraLoc: Repository-Level Fault Localization via Planned Code Inspection

Sangeeta Warrier, Anush Sankaran, Ryen White — Mon, 26 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Agents 42%, Evaluation 31%

A bug report describes what went wrong, but rarely where the code should change. Existing repository-level localization methods often let language models iteratively decide what code to inspect next, requiring repeated model calls and substantial token usage. We investigate whether these inspection decisions can instead be planned before source retrieval. In this research, we introduce CuraLoc, which first selects candidate files and represents their symbols using compact behavioral and structural descriptions. In a single planning decision, the model jointly selects starting symbols, chooses code relationships to inspect from each, and forms hypotheses about likely edits. Deterministic retrieval collects the requested source for a final ranking informed by those hypotheses, concentrating model reasoning on choosing and comparing code rather than directing each retrieval step. We demonstrate the efficacy of our proposed approach on the SWE-Bench Lite and Verified datasets. On SWE Bench Verified with GPT-5.2, CuraLoc raises Func@1 from LocAgent's 53.8% to 59.3% while using 90.3% fewer recorded tokens, and also outperforms RepoSearcher (56.2%) and Agentless (39.6%) on Func@1. On SWE-bench Lite, CuraLoc achieves a high 83.8% Func@5 accuracy improving on Agentless by more than 15% in relative accuracy. Additionally, CuraLoc supports different model backbones, with strong localization accuracy demonstrated using both GPT-5.2 and GPT-6 Astra. These results provide evidence that repository-level code inspection can be planned effectively without iterative model-controlled navigation.

Permission to Retry: A Study of Authorization and Recovery in AI Agent Workflows

Deekshith Marla, Kushagra Bhatnagar — Mon, 26 Oct 2026 00:00:00 +0000

93% · Agents 93%, Evaluation 71%

A durable agent can resume a consequential action while its authority changes and the action's external outcome remains uncertain. We study execution-time reliability of tool-using AI agents, isolating recovery decisions from language-model behavior. Checkpoint state, permission to dispatch, evidence of an external effect and permission to reconcile are distinct obligations. We present a finite conformance study that checks these obligations through histories scored against a provider effect ledger hidden from the recovering application. A prospectively frozen campaign executes 336 histories on native LangGraph and Temporal hosts, crossing 14 schedules, three provider contracts and four matched application policies. All 168 paired host cells have identical checked outcomes. Per host, retained-ledger reconciliation passes all 42 histories: 21 checked completions, 15 denials and six unknown outcomes with reservations retained. Cached retry, revalidation before retry and stable-key retry pass 30, 36 and 31 histories respectively. A separate 16-case component study of InfinityOS, an enterprise AI platform developed by Arya.ai, exercises exported authority, worker, recovery and settlement code over repository fakes. The results isolate the roles of current authority, provider contracts and retained observations in recovery, with safety, completion and unresolved obligations reported separately for the declared application profile.

Can Test-Time Memory Evolution Outperform Test-Time Training for Scientific Discovery?

Yepeng Liu, Shengmai Chen, Xuandong Zhao, Chengzhi Liu, Dawn Song, Xin Eric Wang — Mon, 26 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 61%, New model 51%

Recent \emph{test-time training} (TTT) frameworks have enabled open-weight large language models (LLMs) to outperform inference-only frameworks powered by frontier models on scientific discovery tasks. However, TTT introduces additional computational cost for online weight updates during the discovery process, making computational efficiency an important consideration when scaling to larger models. This motivates us to explore an efficient \emph{test-time adaptation} method for scientific discovery. We introduce \emph{Test-Time Memory Evolution} (TTME), a plug-and-play test-time adaptation method, as an alternative to TTT. TTME continually updates the external memory by distilling the experience to guide subsequent discovery without updating model weights. To have a clean comparison, we evaluate TTME and TTT under the same discovery framework, replacing only the test-time adaptation method. Our evaluation shows that TTME outperforms TTT on all five mathematical tasks, while reducing computational cost by $82.8$-$84.6\%$ in FLOPs and $47.7$-$58.4\%$ in runtime. These results establish TTME as an effective and computationally efficient alternative to TTT in the evaluated discovery settings.

ALLMTS: Automated LLM-based Medical Triage System

Mon, 26 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, Hallucination 48%

Telephone triage is a safety critical task: from a brief intake note, a nurse must decide whether a caller needs an ambulance, an emergency department, an appointment, or self care at home. Queried directly, large language models (LLMs) are unreliable: they over-escalate routine cases and cannot justify their recommendations against an auditable clinical standard. We present the Automated LLM-based Medical Triage System (ALLMTS), which grounds every disposition in the same commercially maintained telehealth guidelines nurses use. Rather than asking an LLM to triage, ALLMTS makes it execute the guideline: it selects the appropriate guideline and evaluates its triage assessment questions. The disposition follows deterministically from the first positive answer, using the same traversal rule a nurse follows. We evaluate ALLMTS on 566 synthetic, expert-validated scenarios spanning 14 dispositions and designed to test specific guideline questions. ALLMTS achieves 0.95 accuracy, compared with 0.47 for a raw-LLM baseline. On a curated 50-scenario subset released as an open benchmark, ALLMTS reaches 0.98 accuracy, on par with the 0.944 correct disposition rate that a five nurse expert panel recorded on those same scenarios. Critically, on that subset the baselines have over-triage rates of up to 0.50 (0.44 on the full dataset), whereas ALLMTS reduces the over-triage rate to at most 0.02 while maintaining low under-triage. These results support coupling LLM language understanding with executable clinical guidelines for triage from intake notes.

LLaCO: Continuous Diffusion for Large Language Models with Competition-Aware Semantic Scheduling

Zijian Jia, Peixuan Zhang, Genying Li, Di Wu, Yujie Fu, Si Li — Mon, 26 Oct 2026 00:00:00 +0000

93% · New model 93%

Continuous diffusion models have emerged as a promising non-autoregressive paradigm for text generation over continuous embedding spaces. Along the continuous trajectory, the residual token identity directly shapes the denoising objective. However, the geometric trajectory misaligns with temporal progression, triggering severe temporal compression. In this paper, we show that embedding norm decay causes Euclidean distance to yield spurious overtaking, indicating that token competition must be evaluated directionally. In this space, a candidate overtakes the source token only when the effective noise magnitude surpasses their semantic separation. We formalize these pairwise dynamics as the Competition Threshold Theory (CTT). Crucially, high-dimensional overtaking requires extreme noise levels, causing standard trajectories to concentrate identity decay into a narrow window near the noisy endpoint. To resolve this, we propose Semantic Retention Scheduling (SRS) to align time allocation with the progression of identity decay, and demonstrate that existing token identity metrics are fundamentally statistical aggregations of these pairwise crossings. Building on these theoretical insights, we introduce LLaCO, a generative framework powered by competition-aware semantic scheduling. LLaCO adapts a pretrained backbone to continuous representations via an embedding directional adapter and timestep-conditioned modulation. To operate within the directional space, LLaCO employs a spherical noising process formulated as a geodesic interpolation toward a mask-biased endpoint, eliminating norm decay while preserving accurate token competition. Guided by SRS, LLaCO reallocates the denoising trajectory proportionally to the logarithmic rank decay. Instantiated as an 8B model, LLaCO outperforms previous continuous diffusion baselines and rivals discrete diffusion models of comparable scale.

On-Policy Self-Training: Data-Free Self-Improvement in Large Language Models

Mon, 26 Oct 2026 00:00:00 +0000

93% · New model 93%

On-policy self-distillation enables post-training foundation models without relying on stronger external teachers. However, existing approaches rely on ground-truth question-answer pairs, human-curated problem sets, or other supervised signals. In this work, we reformulate on-policy self-distillation as a self-training framework, driven by self-generated question-answer pairs and privileged reasoning demonstrations. We push supervision to the limit: guided solely by coarse, high-level topic descriptions, the model synthesizes its own curriculum of queries and privileged reasoning traces without access to any human-curated prompts or answers. We find that self-distillation on fully synthetic privileged information enables a model to improve across multiple domains while retaining its capabilities on domains it is not self-trained on. We demonstrate the generality of this framework across text and multimodal domains, achieving consistent self-improvement on standard language benchmarks and Visual Question Answering. Our findings demonstrate that foundation models possess rich latent capabilities that can be amplified without external task supervision or curated problem distributions.

Interpretable sparse fine-tuning with TopKLoRA enables faithful circuit discovery

Marek Masiak, Andrzej Szablewski, Nicola Cancedda, Constantin Venhoff — Mon, 26 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 52%

Fine-tuning on datasets of unknown origin can create a backdoor in the trained model when the data includes poisoned examples. Difficult to localise and remove, backdoors can persist through safety training. While parameter-efficient fine-tuning methods capture the full parameter changes introduced during training, they are not easily interpretable. We present the full method of TopKLoRA, extending the preliminary proposal of Masiak et al. (2025), an interpretable-by-design LoRA variant in which only the $k$ largest adapter latents are active at each token. Unlike SAEs, TopKLoRA is trained purely on downstream loss, where optimisation pressure incentivises learning functionally causal latents. We demonstrate this sparse LoRA adapter in a sleeper-agent backdoor case study where we create model organisms by fine-tuning gemma-2-2b, Qwen2.5-1.5B, and Qwen2.5-7B on a poisoned instruction-following dataset. For each organism, we identify and ablate a circuit that is both necessary and sufficient to implement the backdoor. Furthermore, our new fine-tuning method produces sparse circuits that are up to an order of magnitude smaller than a dense LoRA baseline, allowing surgical ablation of the malicious capability without sacrificing instruction-following performance. Finally, we propose a technique to construct model organisms with localised circuits using TopKLoRA latents and gradient-routing, which we use to validate the findings of our circuit search algorithm. We consider these findings as a step toward interpretable-by-design architectures that provide faithful units of analysis without compromising model performance. We share our code: https://github.com/interpretable-finetuning/TopKLoRA.

Do "True" and "False" Add Up to One? Complement Consistency in LLM-Agent Opinion Simulations

Anouar Imel, Jin Qi — Mon, 26 Oct 2026 00:00:00 +0000

93% · Evaluation 93%

In LLM-agent opinion simulations, each agent reports the probability that a claim is true after seeing its own previous probability and the probabilities reported by its neighbors, and that number becomes its state in the next round. If the number describes the agent's position rather than the sentence it was shown, then rewriting the claim as its complement and replacing every probability p with 1 - p should return the complementary answer, so the two reports should sum to one. We test this with three small open models: Qwen3-4B, Llama-3.2-3B and Gemma-3-4B. On 2,304 agent states taken from a running network simulation, Qwen's paired reports sum to 1.18 on average, and on 48 real fact-checking claims from AVeriTeC they again exceed one. A controlled benchmark of 24 worlds, each with exactly two possible outcomes, shows the same failure in all three models. The discrepancy is two to ten times larger than the change produced by a same-meaning paraphrase, and it persists under repeated sampling and across peer group sizes and report strengths. When the reports are fed back into two populations that start from complementary states, the populations do not remain mirror images: their divergence is positive after five rounds for every model. Averaging each answer with the complement of its paired answer restores consistency by construction and lowers Brier loss averaged over the two wordings, but at the final round it beats the better of the two original wordings for only one of the three models. A schedule-by-targeting experiment on the same model gives a small, stable interaction because such sweeps keep the claim wording fixed. Consistency under rewording therefore needs its own test before simulated probabilities are read as beliefs

SemantEq: Faithful Bidirectional Equivalence of Formal Statements

Mon, 26 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, Hallucination 51%

Statement autoformalization translates mathematics from natural language into a proof assistant such as Lean 4, and models now do it at scale. As provers get stronger, the burden of correctness shifts from the proof to the statement, that is, to whether the formal statement faithfully translates its source. Given a trusted reference formalization, the natural test is logical equivalence, and it proves too much. $P \leftrightarrow Q$ holds trivially for any two statements $P$ and $Q$ that the library proves, because a prover can prove each separately and pass through $P \leftrightarrow \mathrm{True} \leftrightarrow Q$. The stronger the prover, the more unrelated pairs pass. Earlier methods restrict the prover and check that each proof names the other statement. We introduce SemantEq, a more principled relation that instead requires each statement to turn into the other only through a chain of allowed steps, such as logical rearrangement and rewriting with the library's own equations. Our programs judge a proof by the proof term Lean checks, not by its tactic script. SemantEq-Eval, a search without a model, reaches 100% precision and 65.6% recall on the ProofNetVerif test split under our adjudicated labels, against 95.8% and 61.8% for BEq+, and an 8.9-fold increase in its CPU time adds at most three pairs. A language model finds more proofs, but under two of four prompts it proves both directions of about two in five of the non-equivalent pairs. SemantEq-ProofAudit audits each model-written proof under a policy derived from the relation and accepts none of the 104 non-equivalent pairs, and it raises recall on 407 equivalent pairs from 65.4% to 81.1% under the best prompt. Separating the generation of equivalence proofs from their acceptance lets a stronger prover raise recall without changing what an acceptance means. We release code and data at: https://github.com/hlxr89/SemantEq

Learning Structured Sparsity for Large Models: Bridging Model Compressibility and Efficient Inference

Tianxiang Xia, Kai Lion, Niao He, Michael Muehlebach, Melanie Zeilinger, Hao Ma — Mon, 26 Oct 2026 00:00:00 +0000

92% · Efficiency 92%, New model 85%, Evaluation 30%

Scaling laws continue to reward increases in model scale, while the resulting computational and memory demands drive greater reliance on cloud infrastructure for inference. Yet unreliable network connectivity, communication delays, and privacy requirements can make such reliance impractical, motivating local execution on edge devices. Local deployment requires models to fit within limited memory and execute efficiently with the available computational resources. Recent work has shown that learning sparse-plus-low-rank representations enables elastic compression of a single checkpoint to multiple memory budgets. However, fewer parameters do not necessarily translate into faster inference. Indexing overhead and irregular memory access in unstructured sparse components can offset computational savings. To address this critical deployment barrier, we introduce S-SALAAD, a unified framework that jointly learns the model weights and their low-rank and structured sparse components during training. The framework supports a family of hardware-compatible structured sparsity patterns. It couples this structure learning with budget-aware compression and dedicated CPU and GPU kernels that exploit the learned structure for acceleration, enabling efficient inference from a single checkpoint across memory budgets. Experiments on Llama2 and Qwen3 models, covering both pretraining and distillation, show that S-SALAAD advances the empirical Pareto frontier in perplexity, memory, and inference latency across multiple hardware platforms relative to the evaluated baselines. These results show that structured learning can translate elastic compressibility into reduced memory requirements and lower inference latency, enabling practical deployment of large models on edge devices.

LazyFlash: Forward-Dense, Backward-Sparse Attention for Accelerated Long-Context Fine-Tuning

Keqi Deng, Ruchao Fan, Jinyu Li — Mon, 26 Oct 2026 00:00:00 +0000

92% · New model 92%

As large language models (LLMs) process increasingly long sequences, attention becomes a major computational bottleneck. Sparse attention mostly targets inference, yet many methods still train densely or cannot support sparse training, and their added complexity makes sparsity hard to turn into wall-clock speedups, leaving dense attention widely used in practice. This paper introduces LazyFlash, a forward-dense, backward-sparse attention method built on the observation that attention training is dominated by the backward pass. LazyFlash keeps the forward pass exactly dense, so the trained model remains a standard dense model that existing systems such as vLLM can serve without any change. For the backward pass, LazyFlash proposes a gradient-aware block sparsification design that ranks query blocks by their attention-output gradient norms and skips those contributing negligibly to parameter updates, turning the dense backward into a block-sparse one. A gradient-mass budget adapts the amount of skipped computation to each head and training step, while importance sampling preserves the dense gradient in expectation. These designs are implemented inside the tiled FlashAttention backward kernel, where skipped blocks are never loaded or computed, translating gradient sparsity into real wall-clock acceleration. Experiments on long-context text and speech LLMs show that LazyFlash maintains downstream performance while achieving up to 5.7x attention-backward speedup over FlashAttention and 1.7x end-to-end training speedup.

$M^3$: A Multi-LLM System for Multi-Turn Jailbreaking as a Markov Decision Process

Yuqing Huang, Jack Ji, Jithin Pradeep, Sagarika Suresh, Jekaterina Novikova — Mon, 26 Oct 2026 00:00:00 +0000

92% · Evaluation 92%

Modern Large Language Models (LLMs) remain vulnerable to carefully orchestrated multi-turn interactions that progressively bypass safety mechanisms. However, many existing red-teaming approaches rely on heuristics or impose strong assumptions that limit their applicability in fully black box settings. We address this gap by introducing $M^3$, an automated red-teaming framework that formalizes multi-turn jailbreaking as a finite-horizon Markov Decision Process (MDP). In $M^3$, an attacker LLM interacts with a target LLM while operating over a compact internal state produced by a state LLM. Across two benchmarks and six target models, $M^3$ consistently outperforms state-of-the-art multi-turn red-teaming methods in Attack Success Rates (ASR), achieving improvements of up to 53% in ASR@5. Even with lightweight attacker, $M^3$ attains gains up to 26% in ASR@1 and 34% in ASR@5, compared to baselines that use higher capacity attackers. Beyond improved effectiveness, $M^3$ exhibits steeper risk escalation over conversational turns and achieves more consistent performance across risk categories (up to 4.82 improvement in Gini index) without relying on predefined risk taxonomy, providing a scalable and principled framework for multi-turn red-teaming and safety evaluation.

The LLENS Platform for On-Premises Large Language Model Agents in a Japanese University Hospital: Development and Evaluation Study

Ken Enda, Jien Saito, Masami Mukai, Feng Han, Emir Turkes, Tatsumi Kusakabe — Mon, 26 Oct 2026 00:00:00 +0000

92% · Agents 92%, Evaluation 74%, Hallucination 56%

Large language model agents that use tools could take on part of the electronic health record and desk work that occupies physicians' time, but in many hospitals the medical information systems run inside closed networks, so the model and the agent harness must also sit inside. We designed, implemented, and deployed LLENS, a clinical support AI agent platform that runs inside the closed network of Hokkaido University Hospital on one host with 8 NVIDIA H200 GPUs, the open-weight model Kimi K2.6 (~1 trillion parameters), and a web application rebuilt from Open WebUI. The harness has 10 components, including page image reading, full-text attachment input, a path-addressed document tree without a vector index, a network-isolated sandbox, artifact return, self-checking of generated output, subagents, and saved procedures. Kimi K2.6 was chosen from 3 open-weight candidates and scored 480/500 on the 2026 Japanese Medical Licensing Examination; it generated 119 tokens/s for one user and served 16 concurrent requests at 908 tokens/s in total. On 180 questions over hospital documents (3 seeds each), mean accuracy was 10.7% with no documents, 81.3% with fixed retrieval (top 10 chunks of one lexical search), and 97.2% with interactive access, in which the model reads with document tools and the sandbox (15.9 percentage points above fixed retrieval, 95% CI 10.9-21.3); the difference lay in filtering/counting questions (16.7% vs 87.8%). Five scenarios with synthetic data ran on combinations of the components: needlestick injury response, batch adverse event grading, scanned referral letter transcription, falls data aggregation and plotting, and investigation of 3 topics by subagents. The platform has run on the hospital network since September 14, 2026.

Training an MoE LLM in MXFP4: How We Closed the Gap to BF16

Han Wang, Zhitao Wang, Natalia Frumkin, Xuanwu Yin, Alireza Khodamoradi, Dong Li — Mon, 26 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 48%

Training large language models (LLMs) in sub-8-bit arithmetic promises substantial gains in memory bandwidth and compute throughput, yet FP4 formats introduce quantization error severe enough to destabilize optimization if left unaddressed. We present ALTO, an open-source training recipe for GPT-OSS-20B, a 20-billion-parameter mixture-of-experts (MoE) model, on the MLPerf Small MoE Training benchmark under the MXFP4 microscaling format. The recipe combines established techniques from Abecassis, Felix et al. [arXiv:2509.25149] (hybrid 1D/2D block quantization, randomized Hadamard transforms, and stochastic rounding) with a simplified weight de-oscillation scheme adapted from TetraJet-v2 [arXiv:2510.27527]. On C4 validation at global batch size 16, MXFP4 + RHT + SR + de-oscillation reaches a validation loss of 3.3350 at 16,128 steps, only 0.007 above the BF16 baseline (3.3283), closing roughly 50% of the gap left by MXFP4 without de-oscillation (3.3418). Measured by steps to reach the MLPerf quality target (validation loss 3.34), de-oscillation cuts the FP4 convergence overhead relative to BF16 from +20.0% to +5.0%. We also report negative end-to-end results from differential gradient estimation, outlier clipping, and macro-block scaling, despite each showing operator-level SNR gains. All experiments use simulated MXFP4 kernels on AMD MI300 hardware and wall-clock speed has not been optimized.

DealsFlow: An Evidence-Gated Agentic Harness for Venture Capital Deal Triage

Ming Ze Tang, David Cutting, Boyd Carson, Vasiliki Carson — Mon, 26 Oct 2026 00:00:00 +0000

92% · Hallucination 92%, Evaluation 76%, Agents 70%

Large language models (LLMs) can read startup pitch decks, but letting them emit investment scores directly invites score inflation, run-to-run variance and rationales that are hard to audit. We describe DealsFlow, a human-in-the-loop agentic harness deployed at a UK early-stage fund. The LLM reads the deck, cites evidence and rates each category; code combines the ratings, caps them and only ever lowers them. Three mechanisms implement this: a stage-aware corroboration ceiling (78, 70 or 60 out of 100) on factual categories without external evidence; a sceptic call that may lower but never raise a score; and a pre-fetch pattern that retrieves and indexes all evidence before a single generation pass. We evaluate on $N=243$ SEIS/EIS investments with recorded outcomes. Because recent investments have unresolved outcomes, we use the 142 that are at least three years old. There, the full pipeline's score has Spearman $\rho=0.45$ (95% CI 0.32 to 0.57) with the total-value multiple and AUC 0.71 for identifying write-offs, against $\rho=0.38$ and AUC 0.68 for an unconstrained LLM. The gain is small and comes mostly from the sceptic call; we did not test whether its lower-only rule matters. Investments that later exceeded $2\times$ had 64% of factual categories corroborated, against 38% for write-offs (omnibus Kruskal-Wallis test over five outcome groups, $p=0.031$), and our retrieval was not date-restricted, so these figures may overstate screening skill. In a study of 20 inbound applications, analyst time per application fell from 44.7 to 18.9 minutes including verification and correction, and 84% of 240 cited claims were fully supported by their sources, but 70% of reports needed at least one correction.

Direction-Aware Token Credit Assignment with Reward-Model Gradients

Yikuan Hu, Jifeng Zhu, Minle Su, Jiaqin Yue, Gujie Shao, Yanzhe Chen — Mon, 26 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 34%

GRPO shares a response-level advantage across tokens, leaving unresolved how strongly each token should contribute to learning. Updates induced by the shared advantage reinforce or suppress the sampled token at each position, but these updates do not always point toward increasing reward at the token level. This motivates a work-based view of credit: just as work depends on the component of force along a displacement, token credit should reflect the local reward change along the policy update direction. We introduce Direction-Aware Token Credit Assignment (DATC), which maps frozen reward-model gradients into policy logit space and allocates weights by evaluating the alignment between the reward direction and GRPO's signed sampled-action update direction, while preserving the sign and mean of the response-level advantage. DATC outperforms GRPO in all four settings across two model families.

Speculative Diffusion is a Tale of Two Tradeoffs

Aaron Zorel, Guy Van den Broeck, Aditya Grover, Daniel Israel — Mon, 26 Oct 2026 00:00:00 +0000

91% · Efficiency 91%

Speculative decoding is a technique that accelerates the inference speed of autoregressive large language models (LLMs) by sampling from a faster, lower quality draft model. Traditionally, a fundamental characteristic of speculative decoding is that it is lossless with respect to the verifier distribution, guaranteeing speed without sacrificing quality. This differs from recently developed diffusion LLMs (dLLMs), which are notably able to achieve substantial speedups by degrading the quality of their outputs with less diffusion effort. Because diffusion models are able to sample faster than autoregressive models by generating tokens in parallel, they are considered strong candidates for draft models in speculative diffusion. While it is widely understood that diffusion models can sacrifice quality for speed, speculative decoding itself can also be generalized to generate lossy samples faster using more lenient acceptance criteria with respect to the verifier distribution. Presented with the two parameters of diffusion effort and speculative decoding leniency, this paper explores the full achievable Pareto frontier via joint variations of these parameters. Our experiments show that jointly varying diffusion effort and acceptance leniency can extend the speed-quality Pareto frontier over single step diffusion drafting.

Machine-Readable Operator Manuals over the ISO 11783 Network for Offline, Multilingual LLM Assistance

Timo Oksanen, Julius Gun — Mon, 26 Oct 2026 00:00:00 +0000

91% · Evaluation 91%, Hallucination 32%

If agricultural machinery is to support intelligent, offline assistance, then the system must access technical documentation dynamically. In this paper, the Semantic Synchronization of operator manuals over ISO 11783-based CAN buses is analyzed, defined as the process that brings the local knowledge base of a tractor-side assistant into agreement with the operator manual of the connected implement. Data transfer performance was benchmarked using Markdown, JSON, and XML; various open-weight Large Language Models (LLMs) were also evaluated in cross-lingual query tasks. The automatic evaluation of the answers was validated against four human raters. A complete manual was transferred over a physical ISO 11783 network in 31 to 89 s, depending on format and transport settings, which is comparable to the object pool upload that operators already wait for at the first coupling of an implement. Markdown produced the smallest files, while the question-answering performance was comparable across the three formats for most models. With retrieval-augmented generation, locally run models answered most questions on the English manual correctly, with no advantage of the largest model over 7B and 8B models, but performed worse on the five other language editions. Keyword retrieval failed across languages, whereas semantic retrieval provided the cross-lingual capability. The exchange requires only a new parameter group that uses the existing request and transport mechanisms of ISO 11783, which makes it a candidate for standardization. The study is a proof of concept with inference on PC-grade hardware; deployment on embedded agricultural hardware remains future work.

HyScale: Scaling Long-Context Training for Learned Sparse Attention

Yangsheng Deng, Xinjing Huang, Yulong Wang, Yunzhuo Liu, Jun Hu, Shuyong Tang — Mon, 26 Oct 2026 00:00:00 +0000

91% · New model 91%, Efficiency 44%

We present HyScale, a long-context training system for learned sparse attention. Learned sparse attention is a promising architecture for million-token contexts, widely adopted in production since DeepSeek-V3.2. Selection, however, grows into the training bottleneck. For each query, it scores every previous token whereas sparse attention reads only $k$; worse, current systems materialize a full score table and repeat selection during recomputation. Our key insight is that the output of selection is not an activation, but rather a set of indices. Building on it, HyScale introduces HyTopK and selection replay. HyTopK finds each query's exact top-$k$ on the fly without a score table, while selection replay reuses forward-pass selection in recomputation. Together with holistic system optimizations tackling end-to-end training challenges, HyScale speeds up GLM-5.1 training on 1M-token sequences by 2.07$\times$ over Megatron-LM on 512 H20 GPUs. It also speeds up the CSA layers of DeepSeek-V4-Flash and the QSA layers of Qwen3.8-Flash-Next by 2.01$\times$ and 1.57$\times$ at 1M tokens on 32 H800 GPUs.

Semantically Rich, Metrically Affect-Blind: A Geometric Blind Spot in Modern Text Embedders

Mahdi Naser Moghadasi — Mon, 26 Oct 2026 00:00:00 +0000

91% · Evaluation 91%, New model 75%

We report a systematic dissociation in modern text embedders: sentiment is almost perfectly decodable from their embeddings, yet almost entirely absent from their metric. Across eleven frozen encoders spanning 2021 to 2026, including 2026-generation models and the 4B-parameter LLM embedder Qwen3-Embedding-4B, a linear probe recovers sentence polarity at up to 91.3% accuracy while cosine similarity separates polarity by a margin of only 0.003-0.064. Qwen3-Embedding-4B, the strongest probe in the panel, reaches 91.3% probe accuracy but only 64.3 retrieval MAP. Because retrieval, clustering, deduplication, and retrieval-augmented generation operate on the metric rather than on a trained probe, these systems are effectively blind to information their encoder demonstrably contains. A task-conditioning control suggests this reflects the pretraining objective rather than a prompting artifact: conditioning Jina-v5 on its classification task rather than its retrieval task recovers only about half of the MAP gain that injection achieves on the same weights (63.2 to 73.8, versus 85.1), and leaves the alignment gap essentially unchanged. Sentiment-Injected Sentence Embeddings (SISE) train a small (0.13M-0.69M-parameter) supervised-contrastive head on a frozen encoder in seconds on a CPU, using the labelled training set. The head improves every encoder released since 2022 on full-data probe accuracy (up to +5.8 points), retrieval MAP (up to +26.0) and polarity margin (7-61x). This is a specialization, not a free gain: the compact 128-d head loses 38-41 Spearman points on STS-B, while a fused variant that concatenates it with the base embedding loses 2.4-12.0. With the head, a 0.6B encoder at 128 dimensions matches a 4B encoder given the same head (91.8 vs. 90.3 MAP). We release a CPU-reproducible pipeline and a reusable diagnostic, the Attribute Alignment Gap.

Trial by Trace: Benchmarking AI Agent Harnesses for Reliability and Failure Attribution

Zhiyong Cao, Hengjie Zhu, Yongxin Wang, Xuchen Li, Jintao Li, Shuxia Liu — Mon, 26 Oct 2026 00:00:00 +0000

91% · Evaluation 91%, Agents 71%

LLM agents are increasingly used to solve real-world problems, and their effectiveness depends on both the underlying models and the harnesses that support their operation. However, most existing benchmarks focus on model capabilities and overall agent performance, with limited attention to the evaluation of agent harnesses. To address this gap, we introduce TraceTrial-Bench, which organizes harness reliability into seven components covering context, execution, governance, lifecycle, observability, tools, and verification. We construct 95 expert-level, mechanism-oriented tasks with executable or state-based ground truth to support semantically equivalent comparisons while preserving the native behavior of each harness. We design a Trajectory Tracing (TT) module to capture agent execution traces and introduce a two-layer evaluation framework that separately assesses agent performance and harness reliability. In addition, we design a Multi-source Evidence Acquisition (MEA) module to provide evidence from the agent context, harness, and execution environment for LLM-based evaluation, and develop a Failure Attribution (FA) module that uses explicit mechanism chains to support attribution at the run level. Across 4 harnesses, 9 models, and 95 tasks, our evaluation of 3,420 runs yields macro-averaged harness scores ranging from 59.8% to 72.0%, while 31.1% of runs achieve task success despite failing at least one harness criterion. TraceTrial-Bench provides an auditable and reproducible foundation for comparing agent harnesses, diagnosing their failures, and studying how harness design interacts with model behavior.

Frontier-OPD: Discovering the Edge of Agent Capability and Teaching Agents to Go Beyond

Zheng Jiang, Yiming Chen, Houde Qian, Ling Li, Chaoyang Li, Lifeng Sun — Mon, 26 Oct 2026 00:00:00 +0000

91% · Agents 91%, New model 89%, Evaluation 47%

On-policy distillation aligns teacher feedback with a student's current behavior, yet acquiring useful supervision for multi-turn agents requires identifying both where guidance is needed and whether the student can sustain the interaction afterward. We introduce Frontier-OPD, which addresses this problem through the student's learning frontier: decision situations that remain challenging but admit useful autonomous continuation after brief teacher guidance. To operationalize this frontier, Frontier-OPD organizes on-policy experience into a shared-state trajectory graph, where trajectory-balanced aggregation of teacher-student disagreement identifies distinct intervention candidates while preserving the histories needed for replay. Short teacher bridges are then assessed by comparing agreement over the bridge and the ensuing student continuation with an equally long segment of the original trajectory, providing a behavioral proxy for guidance that the current student can build on. Accepted teacher responses augment ordinary distillation data, and fresh rollouts renew the frontier estimate as the policy changes. Experiments on ALFWorld, WebShop, and ScienceWorld report higher mean autonomous performance than the compared distillation baselines across the evaluated student scales and teacher configurations, with ablations indicating complementary effects of candidate selection and guidance filtering. In one ALFWorld timing setting, the additional collection cost is offset by a shorter time to a target success rate.

FilmFlow: Agentic System for 3D Virtual Film Cinematography

Yiming Zhang, Jiarui Guo, Haojia Wei, Yifei Liu, Yifan Yang, Hongjie Zhang — Mon, 26 Oct 2026 00:00:00 +0000

91% · Evaluation 91%

Complex camera motion is central to visual storytelling, yet translating cinematographic intent into controllable, executable camera trajectories remains challenging. We introduce FilmFlow, an agentic framework that converts natural-language instructions or reference videos into explicit camera trajectories for dynamic 3D environments. To generate these trajectories, a Director agent constructs executable camera plans by composing high-level motions from concurrent atomic operations that jointly control translation, rotation, and framing. The Director is supported by three specialized agents: an Interpreter that extracts scene-grounded cinematographic intent, a Scout that selects initial viewpoints, and a Critic that evaluates rendered results and provides visual feedback for iterative plan refinement. We also introduce CinemaBench, comprising 150 Blender scenes from self-created and licensed sources, with five camera instructions per scene for 750 text-directed tasks and 60 reference videos associated with a subset of scenes. CinemaBench supports evaluation of both text-directed camera generation and video reference-directed camera generation. Across a diverse set of evaluated baselines, FilmFlow achieves the strongest benchmark performance and the highest user preference, demonstrating effective spatial reasoning, temporal coordination, and visual composition for executable 3D cinematography.

From Target Selection to Condition Reuse: Auditing Residual Interventions in Tool-Use Models

Tianda Sun, Dimitar Kazakov — Mon, 26 Oct 2026 00:00:00 +0000

90% · Agents 90%, Evaluation 77%

Activation interventions can make a language model produce a target answer, but can the model use the intended condition in other computations? We study this question in Qwen3.5-27B on tool-use tasks where time or permissions determine eligibility and selection. Across 32 fresh worlds, two case-specific residual-writing variants raise familiar target-selection correctness from 35/128 to 119/128 and from 33/128 to 114/128. Yet these gains disappear when familiar facts and identifiers are recombined within the same selection task, while selection of previously rewarded identifiers increases. The gains persist across output formats but yield little improvement in independent judgments about whether a candidate is eligible or should be selected. Internal edits can nevertheless change how a source fact affects a judgment: in a prespecified secondary comparison, an early-layer source patch changes 50/64 judgments when that source is used, with one adverse change among 64 cases where it is not. This selective effect does not ensure correct complete actions. Together, these results separate target-selection control, condition-sensitive judgments, and complete-action correctness. Evaluating how an edited state is used, rather than only its trained answer, exposes these differences.

Nested Sparsity for Scalable Pretraining and Continual Learning

Tianzhe Chu, Druv Pai, Yi Ma — Mon, 26 Oct 2026 00:00:00 +0000

90% · New model 90%, Efficiency 41%

Mixture-of-experts (MoE) transformers, the standard backbone for modern large-scale deep learning, employ conditional computation: only a sparse subset of model weights is used in each forward pass. In MoE transformers, this sparsity arises in the MoE operator, which generalizes the multi-layer perceptron in the classical transformer. In this work, we explore an alternative mechanism for conditional computation that analogously generalizes the attention operator. The sparsity induced by this mechanism naturally nests (i.e., hierarchically composes) with the sparsity induced by the MoE operator. We show that a transformer employing this nested sparsity mechanism, which we call Nestor, performs better than a conventional MoE transformer at equal compute and equal sparsity in language model pretraining. We also show that Nestor is well-suited for continual pretraining: it can be adapted to longer contexts more effectively than conventional MoEs, as well as adapt more efficiently to new domains while forgetting less on old domains. Furthermore, we demonstrate that these benefits persist as model size, dataset size, and context length grow. Our results demonstrate the potential of alternative conditional computation mechanisms for deep neural network architectures. Code and checkpoints are available on GitHub at https://github.com/Ma-Lab-Berkeley/Nestor .

DIAL-S2ST: Learning Latency Control for Simultaneous Speech-to-Speech Translation

Jeongsoo Choi, Junseok Ahn, Joon Son Chung — Mon, 26 Oct 2026 00:00:00 +0000

90% · New model 90%, Efficiency 42%

We propose DIAL-S2ST, a post-training method for learning inference-time latency control in simultaneous speech-to-speech translation models. Our approach reuses pretrained translation capabilities through self-distillation from lookahead predictions, while limiting the influence of large text distillation losses to account for the imitation gap. This provides supervision for earlier emission without paired target data or reward computation. To make the learned timing adjustable at inference, we jointly train a single LoRA adapter across scales with matched lookahead guidance. Experiments on French-to-English translation with Hibiki demonstrate that only 10 hours of source speech enable this adaptation, providing flexible latency control with a favorable quality-latency trade-off.

Advancing SVD-based LLM Compression via Layer-Wise Error Model Search

Mon, 26 Oct 2026 00:00:00 +0000

90% · Efficiency 90%, Evaluation 46%

Low-rank SVD-based compression offers a powerful strategy to reduce the computational costs of LLMs. However, existing methods face two key limitations: (i) global rank allocation, where uncalibrated error proxies fail to capture complex error propagation, and (ii) decomposition quality, where Fisher-based estimators suffer from severe rank collapse. In this work, we address these limitations by introducing Layer-wise Error Modeling Search (LEMS) and KFAC-SVD. LEMS advances rank allocation by introducing a layer-wise error surrogate that integrates local and global layer importance alongside a propagation bias, enabling effective global rank allocation via an ILP formulation. KFAC-SVD improves decomposition quality by utilizing token-wise statistics, mitigating the rank deficiency observed in prior Fisher-based SVD approaches. Across Mistral, Qwen3, and Llama 3 model families, we show that LEMS consistently outperforms existing search strategies, delivering significant zero-shot accuracy gains of up to 4.8 p.p. that generalize to model sizes of 70B parameters, while KFAC-SVD achieves an average perplexity improvement of 15%. Project page and code: https://lems-svd.github.io

Yalda-Embedding: Open Development of a Competitive Persian Retrieval Model

Yalda Taheri, Mohammad Hassan Heydari, Armon Rasooli — Mon, 26 Oct 2026 00:00:00 +0000

90% · New model 90%, Evaluation 51%, Efficiency 41%

Despite advances in multilingual text representations, effective dense retrieval for Persian remains constrained by the limited availability of language-specific training resources and compact embedding models. We present Yalda Embedding, a 595.8-million-parameter Persian retrieval model adapted from Qwen3-Embedding-0.6B using more than one million query-document pairs. On eight Persian retrieval tasks from FaMTEB, the model achieves the highest macro-average among the evaluated 0.6B models, scoring 0.5488 and outperforming Qwen3-Embedding-0.6B and F2LLM-v2-0.6B on every task. It also approaches Qwen3-Embedding-4B, winning on four tasks, tying on one, and trailing by 0.0075 in macro-average. An incremental ablation study examines contrastive batch size, optimization stabilization, and relational knowledge distillation from F2LLM-v2-8B. Under the one-epoch training setup, contrastive fine-tuning alone yields a modest mean improvement over initialization and regressions on two tasks. The final distilled configuration improves all eight tasks over the non-distilled model, raising the mean from 0.4813 to 0.5488 without increasing inference-time model size. These gains include hard-negative evaluation tasks despite training without explicitly mined or supplied hard negatives. We release model weights, training and evaluation code, the pairwise dataset, and its teacher-embedded counterpart, enabling inspection of data composition and training procedures, independent evaluation, and reuse without repeating expensive teacher inference.

Agent Mosaic: Recognizing Confidentiality Leaks in Agentic LLM Systems

Daniel Ruiz, Luca Luceri, Emilio Ferrara — Mon, 26 Oct 2026 00:00:00 +0000

90% · Evaluation 90%, Agents 65%

Modern LLM agents can retrieve information from accessible documents, and then disclose a sensitive conclusion assembled from them. This mosaic effect creates a critical risk for agents operating over confidential data. We quantify its realization with 30 synthetic cybersecurity, medical, and military examples, yielding 10,800 final answers from three LLM agents in controlled simulation and 1,440 answers from live agents in two open-source harnesses. Even under benign queries designed to be answerable from one document, GPT-5.6-Sol, the best performing model tested, discloses forbidden composites in 7.08% of answers (95% CI: 3.75-10.67%) in simulation, with worse disclosure rates in live harnesses. Adding visible classification headers and explicit aggregation rules reduces pooled disclosure rates substantially, yet unsafe disclosures persist in 32.37% of simulated answers (95% CI: 28.33-36.19%) across all intents despite these safeguards. Furthermore, increased refusals do not guarantee confidentiality, since significant refusals reveal the prohibited conclusion regardless. These results point to a troubling conclusion: sensitive data leaks via mosaic-style events are an observed confidentiality failure mode in the evaluated settings at non-negligible rates. They motivate stronger controls on what agents may infer and disclose in sensitive data environments with mixed-clearance users.

★ FusionEval: Benchmarking Model Fusion for Large Language Models

Shuo Cai, Yanggan Gu, Yuhang Liu, Su Lu, Yuanyi Wang, Yang Yu — Fri, 23 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Model fusion aims to integrate the capabilities of multi-expert models into a single model. Existing evaluations offer limited coverage of reasoning with long chains of thought and tool use, while strong average performance can mask substantial losses in individual expert capabilities. We introduce FusionEval, a benchmark that provides controlled expert pools, integrates fusion methods with and without additional training, and offers a unified evaluation suite. The pools contain domain experts independently trained from Qwen3 and Llama base models across math, code, science, instruction following, and tool use. The suite evaluates capability retention and computational cost on eight public benchmarks under comparable data access rules, reporting normalized average and worst performance, runtime, and peak GPU memory usage. Extensive experiments show that training-free methods preserve reasoning performance but substantially degrade tool use at both model scales. In contrast, MOPD preserves tool use and achieves a more balanced capability profile, but requires additional data preparation, inference, and training. Jointly retaining reasoning, instruction following, and tool use at low computational cost remains a challenge. We hope FusionEval supports future research on efficient model fusion that jointly preserves expert capabilities. Our code is available at https://github.com/Baicaihaochi/FusionEval_Benchmark.

★ Rethinking the Evaluation of LLM Fact-Checkers

Fri, 23 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Hallucination 78%

Large language models (LLMs) are widely used to answer factual questions, but their answers can contain errors and must be fact-checked claim by claim. Yet the resulting score depends on three design decisions: which questions the claims answer, under what conditions each claim is judged, and which sources count as evidence. No existing benchmark specifies all three. We introduce JSFactCheckBench, a Japanese benchmark that specifies all three, with 1,913 claims from LLM answers to questions about misleading information that circulated on Japanese social media, each annotated for factuality at a reference time and for importance to the question, using evidence from preselected Web sources. Across six LLMs, all verify claims directly answering the question less accurately than supplementary claims, a gap that aggregate scores obscure. Scores also drop when models must locate the relevant content in longer or multiple sources and, for all four open models, when the surrounding question and response are added. These weaknesses are visible only because the benchmark records the conditions behind each label. A benchmark that leaves them implicit cannot reveal them.

★ EpiSEM-Bench: Evaluating Safety and Evidence Recovery in LLM Agents under Missing Evidence

Qijin Zhang, Hanzhen Wang, Yuehan Zhao, Haoyu Dai, Yao Shu, Zhongxiang Dai — Fri, 23 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 96%

Tool-using language model agents rely on environmental facts to decide whether an action is permitted. When a tool response omits a safety-relevant fact, an agent may carry out a prohibited action even though the user request is benign and every returned statement is true. We introduce \bench{}, an executable benchmark for evaluating unsafe execution, safety boundaries, and evidence recovery in LLM agents under missing evidence. It contains 300 matched unsafe/safe task pairs across six domains. By systematically suppressing critical facts, the benchmark measures unsafe execution under critical evidence omission and determines the minimum observed failure cardinality, including failures with full evidence and cases with no observed failure after exhaustive testing. An independent audit tool tests whether agents seek missing facts and then withhold prohibited actions while completing permissible tasks. The benchmark reveals substantial variation across the ten models in omission robustness, evidence acquisition, and appropriate action. Resistance to missing evidence does not necessarily translate into useful recovery. Recovery access reduces unsafe execution for all ten models, but agents differ in whether they actively obtain the missing facts and use them to act appropriately in both worlds.

★ CulturalRiddles: A Multicultural Benchmark of Riddles

Fri, 23 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

As LLMs expand their language coverage, evaluating how well they understand and represent diverse cultures becomes increasingly important. However, many benchmarks remain English-centric, with limited cultural coverage and a focus on factual knowledge over its application in real-world cultural contexts. We propose CulturalRiddles, a benchmark of original, culturally grounded riddles created entirely by humans through a participatory community effort, spanning 62 cultures (45 countries, 51 languages). These riddles are deeply rooted in their respective cultures and require abductive and compositional reasoning, alongside detailed factual and social knowledge, language skills, and an understanding of culturally specific associations. We benchmark 28 LLM configurations using both automatic and large-scale human evaluation. Even frontier models achieve at most 22% accuracy in some communities. We publicly release all artifacts alongside an open contribution process to expand the benchmark's linguistic and cultural coverage.

★ Quant-Arena: Benchmarking LLM Agents for Quantitative Trading Research

Houde Qian, Zheng Jiang, Wenyi Li, Chenxi Xu, Xinjie Lin, Yuzhao Peng — Fri, 23 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Quantitative research is an iterative loop: a researcher proposes a market hypothesis, implements it as an executable signal, tests it on historical data, and decides what to keep. Existing benchmarks score formulas, code, or individual trading decisions, leaving open whether large language models (LLMs) can complete this loop and produce strategies that work beyond the research period. We introduce Quant-Arena, a benchmark and public leaderboard for end-to-end quantitative research on real A-share market data. A model constructs features and stock-scoring code, receives feedback only on 2022--2023, and freezes one strategy for assessment on the sealed year 2024. A cash-account evaluator applies trading constraints and costs, including market impact estimated from a simulated order book. Scores are dominated by realized excess return and complemented by signal quality and the diversity of features used by the strategy. We evaluate nine frontier models, among which GPT-6 Astra scores highest; yet no model beats the equal-weight market in 2024. Quant-Arena is open to submissions from any model or agent harness.

★ TauIndianBankBench: Procedure-Aware, Action-Oriented Evaluation of Tool-Using Conversational Agents

Krishanu Adhikary, Prashant Devadiga — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 84%, Hallucination 31%

Many Indian banks serve customers through FAQ pages and scripted chatbots. Assistants based on language models could respond faster and carry out requests such as blocking a card, but only if they give correct information and follow the bank's rules. These rules prescribe a procedure, a set of required steps, such as obtaining a closure quote before closing a deposit. Adapting models to this domain requires measuring how well they follow the procedure, before and after training. Benchmarks usually judge only the records left at the end of a conversation. Many required steps, however, leave the records unchanged, so an assistant that skips them can still be judged successful. We present TauIndianBankBench, an action-oriented benchmark of 1,000 Indian retail-banking tasks where an assistant helps a simulated customer. Besides the final records, it checks that the assistant took every required step and made no unrequested change. For every model, 6.5-12.0% of sessions are judged successful when only the final records are checked, although the assistant skipped a required step or made an unrequested change. Scoring both procedure and outcome shows more accurately how well language models perform in Indian retail banking.

★ SIM-UT: A User Simulation based Evaluation Framework for Medical Triage Chatbots

Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

LLM-based chat agents are increasingly used for complex, multi-turn, multilingual medical triage in low- and middle-income countries (LMICs), yet evaluation methods focus on single-turn, English-only benchmarks that primarily test the chatbots' knowledge. In practice, users are often confused, overwhelmed, or emotionally vulnerable, giving unclear, misspelled single-word answers or long messages in regional dialects, and the context of a conversation can shift rapidly. We propose SIM-UT (Simulation of User Behaviour for Triage), a multi-turn, simulation-based evaluation method that identifies ecologically valid, contextually relevant failure modes of LLM-based triage chat agents. Using SIM-UT, we evaluated three real-world WhatsApp-based health chatbots: a respiratory screening programme in India, a hearing health support programme in South Africa, and a diabetes and lifestyle coaching agent in Kenya. We compared a medically fine-tuned open model (MedGemma) against general-purpose frontier models (ChatGPT and Claude) across 5,000+ multilingual simulated conversation turns grounded in the personas, constraints, and communication patterns of the real user populations, over 2-4 iterative rounds. A key failure mode emerged in both model families: de-escalation risk. The agents initially identified medical emergencies correctly, but when the user minimised the symptoms without introducing new clinical information, they de-escalated, exhibiting clinically dangerous sycophancy, for example agreeing that vaginal bleeding in pregnancy was "not a problem" or that decreased fetal movements at 32 weeks "can be normal". We recommend user simulation-based behavioural evaluation, grounded in real user context and covering adversarial and edge cases, as standard practice for triage chatbots in production, so that safeguards can be added before deployment or in subsequent iterations.

★ Why Bother Me Again? Benchmarking Personal Agent Memory for Interaction Preferences

Jialin Li, Zhenhao Chen, Hanjun Luo, Hanan Salam — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Personal intelligence systems are taking on increasingly complex everyday tasks, yet repeated questions, unnecessary confirmations, and unwanted outreach can turn capable assistance into interaction burden. Reducing this burden requires agents to remember how users want to be assisted. However, existing memory benchmarks focus on factual recall and personalized answers and rarely test how agents learn and act on interaction preferences. We introduce MemPABench, an interactive environment for on-policy evaluation of memory in personal agents. MemPABench poses long-horizon tasks requiring agents to decide when to clarify, confirm, or act autonomously. Four grounded simulated users hold 14 preferences across work and personal contexts. Their feedback spans 99 sessions over four months, with five event-driven changes per user testing memory updates. We evaluate preference selection, delivered behavior, and sensitivity to context and preference change. Evaluating four memory systems across four LLM backbones, we find that current memory brings limited gains in preference selection. Retrieval-based systems (i.e. Mem0, MemOS, simple RAG) do not outperform No Memory, while File-based memory improves accuracy but applies preferences beyond relevant contexts. Our analysis shows that current memory struggles to retrieve interaction preferences and adapt to preference changes. Even with the correct preference available, overriding the agent's default interaction habits remains challenging. The environment and dataset are available at https://mempabench.com.

★ IFC-MCQ: Automatically Generating Multiple-Choice Benchmarks for Evaluating LLM Agents on IFC Data Retrieval

Rabindra Lamsal, Sisi Zlatanova, Xuesong Shen — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 65%, Hallucination 30%

Industry Foundation Classes (IFC), the standard for building information exchange in architecture, engineering, and construction (AEC), is increasingly queried through large language models (LLMs) using natural language. Evaluating these systems at scale is difficult because free-text answers require costly human or LLM-based review. We propose IFC-MCQ, a method that automatically generates multiple-choice benchmarks from IFC files for exact option-set grading. The taxonomy covers 15 categories and 36 subtypes across four groups, with answer keys derived from the file and distractors constructed from file-derived values or predefined rules. We applied IFC-MCQ to IFC models of varying sizes and modelling characteristics to assess generation coverage and evaluated two agentic configurations on selected models using tool- and code-based retrieval. The evaluation exposed category-specific failure modes and showed that performance can vary with the source IFC model and retrieval interface. IFC-MCQ enables scalable benchmarking of LLM-based IFC retrieval in AEC workflows.

★ PLE: Physical-Engineering's Last Exam

Rujun Han, Yufan Zhuang, Arijit Ray, Han Yu, Zoey CuiZhu, George Lee — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 80%

Frontier language models now score above 90% on popular knowledge benchmarks and are increasingly deployed as agents that carry out multi-hour professional work. Yet no existing benchmark tests whether they can perform the real-world, economically valuable work of a professional physical engineer, such as modelling physics and legal code requirements across a multi-format project corpus, optimising material and installation costs, and issuing 2D/3D engineering design packages that a licensed reviewer would sign off on. Existing engineering benchmarks focus on textbook question answering, isolated tool calls or self-contained toy design problems. As a result, most of these benchmarks are already saturated by leading agents and solving them is far away from delivering real engineering design. We introduce PLE, a benchmark of 85 long-horizon, agentic tasks in professional physical engineering consisting of electrical, civil, structural and mechanical engineering, where solving a task means completing a realistic, billable engineering deliverable. PLE combines two complementary curation paths. Expert-Authored tasks are created end to end by practising engineers from a professional data provider's expert pool. Co-Authored tasks start from problem statements written by practising engineers; an adversarial Creator-Solver agent pipeline constructs each task's realistic project environment, executable grader and reference deliverables before sending them to expert engineers for validation and difficulty intensification. We evaluate a variety of frontier LLMs under seven publicly available agent harnesses. Even the strongest configuration reaches a mean score of only 50.6%, while spending 27M tokens and 63 minutes per task on average, demonstrating the substantial gap between current agents and autonomous physical engineering design.

★ Diagnosing Personalization Failures in Mobile GUI Agents

Zhixin Lin, Dongliang Xu, Jungang Li, Shidong Pan, Jun Feng, Yanbin Sun — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 77%

Unlike conventional GUI agents, personalized GUI agents require user preferences to guide sequential execution over changing GUI states. This raises two challenges: determining HOW each preference should alter execution strategy and WHEN it should take effect along the interaction trajectory. Existing benchmarks primarily evaluate overall task success, offering limited insight into these questions. We introduce PrefDiag-Bench with two components: (1) a categorization of preferences into five types, characterizing HOW they affect execution; and (2) a stage-wise diagnostic framework with 1,664 probes, tracing WHEN preferences should take effect and where failures occur. We evaluate eleven GUI agent systems, including end-to-end agents and agentic workflows, under two forms of user context: explicit user profiles and behavioral memory logs. Although existing agents achieve relatively high completion rates on non-personalized tasks, performance is substantially lower on preference-personalized tasks. Even the best-performing model, Seed-2.0-Pro, achieves only 47.17% on Process preferences under explicit user profiles. Agents struggle particularly with Ranking and Process preferences. Further analysis reveals two key capability bottlenecks: selecting relevant evidence from user history and recognizing when preferences should take effect during execution. Controlled interventions that explicitly provide relevant evidence improve end-to-end task success by up to 7.06%. PrefDiag-Bench offers a more systematic evaluation of GUI agents' personalization capabilities beyond overall task success. Code and data are available at https://github.com/Zhixin-L/PrefDiag-Bench.

★ TutorCaseBench: A Case-Grounded Benchmark for Evaluating LLM Tutors in Supporting Students with Challenging Behaviors

Yeon Su Park, Chanyoung Park, Tae Soo Kim, Seungju Kim, Hyunwoo Kim, Daho Jung — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

LLMs are increasingly used as AI tutors, yet existing benchmarks overlook the challenging behaviors common among students with learning and behavioral difficulties. We introduce TutorCaseBench, a benchmark that grounds AI-tutor evaluation in qualitative case studies from pedagogy, educational psychology, and special education. Our pipeline converts documented student categories into case-specific scenarios, tutor-student dialogues, and rubrics grounded in source evidence. The benchmark contains 923 dialogues across five challenge types, evaluated through open-ended response generation (1,560 items) and multiple-choice selection (2,537 items). Across 11 state-of-the-art LLMs, even the strongest model satisfies a rubric criterion in only 55.8% of open-ended cases, while models are better at selecting appropriate strategies than generating situated responses. These results reveal a gap between intervention recognition and context-specific adaptation, especially for open-weight models. All data are publicly available at https://github.com/kixlab/TutorCaseBench.

★ TRIAD: Benchmarking Omni-Modal Ambiguity in Multimodal Large Language Models

Zhaolu Kang, Yidi Wang, Yile Li, Siqi Zeng, Yingjie He, Richeng Xuan — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Ambiguity is a central feature of natural communication: a sentence, image, or sound may remain underspecified until evidence from the other channels is considered. Yet current multimodal benchmarks mostly test perception, recognition, or reasoning over already determinate inputs, leaving unclear whether omni-modal large language models can resolve ambiguity distributed across text, vision, and audio. We introduce TRIAD, a benchmark for omni-modal ambiguity resolution. Each item is a text--image--audio triplet paired with a question and a set of answer options, and is constructed so that the full triplet determines a unique answer while removing any one modality makes the set of consistent options non-unique. TRIAD contains $327$ hand-written queries in $86$ scene groups and covers $18$ ambiguity categories across the three modalities. Evaluating $14$ omni-modal systems reveals that strong item-level performance does not translate into robust scene-level disambiguation: the best model still trails humans by over $20$ points in group accuracy. Removing answer options widens the gap further, driving model group accuracy to the single digits while humans remain robust. Leave-one-out and transcript/description counterfactuals further show that current systems often underuse audio cues when text and image are present. TRIAD reframes omni-modal evaluation around disambiguation rather than recognition, and provides a taxonomy, benchmark, and diagnostic protocol for measuring this capability.

★ ColaBench: Evaluating Memory-Guided Knowledge Acquisition in Personal Agents

Wenjie Zhang, Zhichao Shi, Xuhui Jiang, Xiaojun Wu, Cehao Yang, Chengjin Xu — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 38%, Hallucination 32%

Personal agents need to draw on long-term interaction history to retrieve and apply external knowledge suited to a user's current request. Even an answer grounded in relevant documents can be unsuitable when it overlooks the user's circumstances. This creates two connected challenges: using historical conditions to identify the knowledge needed for a new request, and applying that knowledge to the user's situation. Existing benchmarks advance memory recall, personalized retrieval, and memory-guided action, while the joint evaluation of these two challenges remains underexplored. We introduce ColaBench, a synthetic benchmark in health and daily living, with 349 tasks across 10 Profiles and a knowledge base of 13,033 documents. Its Evidence Graph connects personal conditions to their disclosures across public interaction sessions, records prior document exposure, and links answer requirements to personal and newly acquired document evidence. We also introduce MemKB, which uses state summaries to inform queries and rule application, and memory-document links to retrieve knowledge associated with the user's history. MemKB achieves 86.25% Accuracy with an average of 4.07 interaction rounds per task. Task-level analysis finds larger gains on tasks integrating information across multiple historical Facts; error analysis further identifies unsupported personal assumptions and rule misapplication as obstacles to using retrieved evidence. ColaBench provides traceable tasks and an evaluation framework for studying how memory guides knowledge acquisition and rule application.

★ IndoNutriQA: An Indonesian Clinical Nutrition Question Answering Benchmark

Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 64%

Clinical Question Answering (QA) underpins Clinical Intelligent Decision Support Systems, but progress is severely limited for low-resource languages such as Indonesian by the absence of public domain-specific data. We introduce IndoNutriQA, an expert-curated Indonesian clinical QA benchmark of 911 question-answer pairs built from real clinical-nutrition reports. Rather than a large-scale training corpus, it is positioned as a high-quality evaluation resource: every pair is grounded in a clinically relevant context, enriched with structured metadata (anthropometric assessment, medical diagnosis, document provenance, chunk segmentation). Eight standardised question types following the Nutrition Care Process (NCP), including the Problem, Etiology, Sign and Symptom (PES) diagnosis statement, are answered for every report, which was written by a graduate-level dietetic intern, by one of three nutritionists. Answers are free-form clinical prose rather than extractive spans, and an expert validation study finds 6-10% of reference answers contestable. The dataset supports generative, extractive and retrieval-augmented QA as well as zero- and few-shot LLM evaluation. We report fine-tuned baselines (IndoBERT variants, mT5, RAG) with 5-fold cross-validation, learning curves and bootstrap 95% confidence intervals, and benchmark eight contemporary LLMs zero- and five-shot. We further calibrate our LLM-as-judge rubric against three expert raters (one RD, two nutritionists), finding it human-comparable on factuality and completeness but unreliable on clinical safety, and a human error analysis shows it misses unsupported content that the raters identify. IndoNutriQA is released to foster research in Indonesian clinical NLP and nutritional decision support.

★ PointerBench: A Diagnostic Benchmark for Fine-Grained GUI Grounding

Valentino Sacco, Saurabh Pandey, Alexander Florian Jercher, Maximilian Schilling — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Multimodal Large Language Models have shown promising performance in computer-use tasks, from planning user workflows to grounding natural-language actions to precise UI targets. Existing end-to-end benchmarks primarily emphasize long-horizon task completion, while GUI grounding benchmarks remain focused on conventional controls and a limited set of professional interfaces. We introduce POINTERBENCH, a diagnostic suite of 1,500 screenshot-instruction pairs probing three complementary capabilities: thin sub-element identification and geometric grounding in spreadsheets (POINTERBENCH-Sheets), character-level grounding in long-text documents (POINTERBENCH-Text), and mixed GUI grounding across 100 professional applications (POINTERBENCH-Pro). We carefully curate each task with pixel-exact target geometry and evaluate both general-purpose multimodal models and 39 specialist GUI grounding configurations. We find text grounding to be the most challenging setting, with Claude Fable 5, GPT-5.6 Sol, and the leading native specialist UI-Venus-2-9B achieving 69.6%, 53.4%, and 25.8% success on POINTERBENCH-Text despite respective 83.2%, 77.7%, and 54.5% average success across POINTERBENCH. These results reveal substantial headroom in fine-grained GUI grounding across frontier and specialist models. We publicly release the benchmark and evaluation code at https://github.com/warmwindOS/pointerbench.

★ When LLMs Forget What They Believe: Characterizing State Integrity Collapse in Long-Horizon Dialogues

Yancheng Zhu, Yifan Yuan — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Large language models can repeat an instruction while failing to keep obeying it. We study this recall-compliance dissociation in long-horizon dialogue and define State Integrity Collapse: a model preserves a stated constraint in memory but stops applying it to later turns. We introduce StateTrack-Bench, a controlled suite of conversations covering numerical, format, role/persona, factual-memory, and logical-rule constraints. Across seven provider-pinned model snapshots and five dialogue-depth probes, strict joint retention ranges from 62.4% to 100.0%, with degradation concentrated in symbolic constraints and in one especially brittle model. A forked recall audit shows that 88.6% of observed failures are behavioral-drift failures rather than ordinary forgetting; excluding the brittle outlier preserves this diagnosis (84.4% of failures). Extended evaluations at depths up to 50 turns confirm that the dissociation persists beyond the core benchmark, while Structured State Anchoring recovers up to 24.4 points on immutable constraints. The results show that reliability cannot be inferred from single-turn instruction following alone; evaluation must track whether remembered constraints remain operational as dialogue state grows. Code and artifacts are available at: https://github.com/xzzvsxd/StateTrack.

★ Parameter-Efficient Adaptation of Large Language Models to a Low-Resource Agglutinative Language: A Corpus, Datasets, and a Reproducible Benchmark for the Bashkir Language

Mullosharaf K. Arabov, Svetlana S. Khaybullina — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Bashkir is a low-resource agglutinative language written in Cyrillic and remains poorly represented in modern LLMs. We present an open resource suite and a systematic benchmark of adaptation methods: the Bashkir Web Corpus (71,567 documents, 46.9M tokens, 16 sources) and three labelled datasets for binary, multiclass and multilabel text classification; the benchmark covers the first two. Experiments use eight architectures (82M-8B: DistilGPT2, GPT-2 base/medium, Phi-3-mini, Mistral-7B, YandexGPT-5-Lite-8B, mBERT, XLM-R) and three adaptation groups (full fine-tuning, QLoRA, in-context learning), plus continual pretraining (CPT). (1) QLoRA on 3.8-8B models outperforms encoder-only fine-tuning on binary classification: YandexGPT-5-Lite-8B with CPT reaches macro-F1 0.748 vs 0.653 (XLM-R) and 0.650 (mBERT); pairwise differences are significant after Holm correction (p<0.001). (2) YandexGPT-5-Lite-8B significantly outperforms Mistral-7B and Phi-3-mini on multiclass classification, BPB and native-speaker ratings (mean 3.28 vs 2.52 and 2.22; n=100; alpha=0.69-0.77; p<0.05); on binary classification its QLoRA advantage is not significant (p>=0.084). (3) CPT effects depend on model and task: perplexity improves for Phi-3-mini (+0.58) and YandexGPT-5-Lite-8B (+0.33) but worsens for Mistral-7B (-1.15, p<0.001); on binary classification CPT helps Mistral-7B most (+4.0 macro-F1). (4) Tokeniser fertility on Bashkir varies 2.67x (2.92-7.80) and tracks adaptation quality, while perplexity across tokenisers is not comparable. Cyrillic tokenisers with Russian pretraining are preferable; LLM judges reproduce native ratings for grammar and coherence (rho=0.33-0.46) but not language correctness (rho=0.16, p=0.11); native evaluation remains necessary. All artefacts are open-licensed.

★ DuoCook: An Overcooked-Inspired Benchmark for Visual Planning and Coordination

Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Coordinating two players in a cooking task requires a vision-language agent to track changing scene states, sequence recipe steps, and allocate shared resources under a time budget. We introduce an Overcooked-inspired benchmark for evaluating these capabilities in Unreal Engine environments constructed using coding agents. A single vision-language model controls both players from top-down visual observations, selecting high-level actions while the engine executes navigation. The benchmark contains 20 maps in five categories. These categories support evaluation of centralized coordination under equipment contention, changing terrain, and handoff dependencies. A prop legend and pot ingredient-count indicators provide explicit visual cues, while grid-based destinations support spatial grounding. The evaluation compares four VLMs on all 20 maps, measuring task score, order completion, delivery precision, action acceptance rate, and model API expenditure. A memory ablation on the same 20 maps examines the effect of a shared, self-authored notebook while retaining six-response action history in both conditions. The strongest evaluated model, \texttt{gpt-6-astra}, achieves 26.2\% mean order completion. Notebook memory has mixed score effects across models, with positive mean differences for Claude and Qwen and negative differences for GPT and Gemini.

★ Deep Search in Real-World Web Environments

Xiaomin Yu, Hongliang Chen, Zhe Huang, Shilin Yan, Shao-Rong Su, Xuqiao Ma — Fri, 23 Oct 2026 00:00:00 +0000

98% · Agents 98%, Evaluation 98%

Despite substantial progress in Deep Search agents, their ability to conduct end-to-end research in real-world open Web environments remains insufficiently evaluated. Unlike static retrieval settings, information on the real-world Web is distributed across diverse sources, modalities, and heterogeneous formats, often involving complex interaction mechanisms that require agents to autonomously search for, access, and integrate relevant evidence during execution. To address this gap, we introduce the OpenDS task, designed to evaluate agents' ability to perform multi-step deep search in the live, open Web. This task requires agents to autonomously explore the real Web, discover and connect relevant evidence from multiple independent sources through Web search, browser interaction, and evidence processing, and ultimately determine a unique factual answer. Building on this task formulation, we construct OpenDS-World, a benchmark comprising 200 challenging tasks spanning diverse information modalities, heterogeneous file formats, and interactive access requirements, with independent human verification to ensure evidence validity and answer uniqueness. We evaluate 12 frontier models under a unified tool environment and execution protocol. The best-performing model achieves 39.0% accuracy, while human evaluation across 1,000 independent attempts yields 21.1% accuracy. Further analyses of tool usage, search cost, termination behavior, and failure recovery reveal substantial differences in research strategies across models, as well as persistent challenges in tool coordination and long-horizon search management. OpenDS-World provides a unified evaluation platform for systematically assessing and advancing agents' autonomous deep search capabilities in real-world open Web environments.

★ Code2Video-Compass: A Comprehensive Benchmark for General-Purpose Code-Based Video Generation across T2V, I2V, and V2V

Zhenghan Chen, Zekai Shao, Xin Lin, Leting Cheng, Siming Chen — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Code-based video generation offers fine-grained control, targeted editing, and component reuse compared with direct video generation. This approach requires multimodal large language models (MLLMs) to translate visual and temporal requirements into executable programs that produce complete videos. However, existing code-based benchmarks primarily target specialized tasks, leaving complete video creation across diverse creative categories and input modalities insufficiently evaluated. We introduce Code2Video-Compass, a general-purpose benchmark for code-based video generation across text-to-video (T2V), image-to-video (I2V), and video-to-video (V2V) settings. Code2Video-Compass comprises 2,104 tasks spanning 10 categories, 24 subtypes, and three complexity dimensions covering construction, dynamic animation, and state dependency. It combines tasks curated from textual instructions and visual references with corresponding T2V, I2V, and V2V tasks derived from complete professional and user-created videos. We propose a dual-level evaluation framework combining 12 objective metrics and five subjective metrics. Objective evaluation covers operational performance, video quality, text-video alignment, and reference fidelity, while subjective evaluation uses an MLLM judge to evaluate instruction following and holistic video quality. Extensive evaluation of 15 recent proprietary and open-weight MLLMs from nine families reveals that even leading models face challenges in reliably generating high-quality complete videos aligned with textual and visual requirements, with performance generally declining as construction, dynamic animation, and state-dependency complexity increases. We anticipate that Code2Video-Compass will advance reliable, coherent, and controllable code-based video generation and inspire broader progress in programmatic visual creation.

★ VerifierBench: Benchmarking the Generation of Executable Verifiers for LLM Agents

Jianing Yin, Ziyi Chen, Hao Wang, Yiyou Sun, Zijian Jin, Dawn Song — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 54%

Agents need reliable ways to validate their outputs, both when completing tasks and when filtering synthetic training data. We study verifier generation: from the instruction and environment available for solving a task, an agent writes an executable verifier that must accept every valid solution and reject invalid ones. We introduce VerifierBench, a benchmark of 222 tasks from six agent benchmarks spanning tool-call, state-based, and code verification, with 4,993 hidden test cases that include alternative valid solutions and attacks; Verifier4Verifier (V4V), a six-check audit, screens its source verifiers and test suites. Across 19 models in a common agent harness, normalized balanced accuracy ranges from 10.6 to 62.1, with generated verifiers consistently rejecting invalid solutions more reliably than they accept valid ones. This apparent strength can mask overly restrictive verification: most false rejections in the analyzed records come from verifiers that accept no tested solution to a task. Case studies reveal unsupported assumptions about solution formats, tool use, and code changes that exclude valid solutions. These errors survive self-testing because agents test solutions that share the verifier's mistaken assumptions, or submit verifiers without confirming that any correct solution passes. These findings suggest testing both whether a verifier accepts a correct solution and whether it accepts other solutions permitted by the task. The project page and dataset are publicly available.

★ EvaOmni: An Evidence-Grounded Benchmark for Audio-Visual Reasoning

Xiaoxing Hu, Wenxi Yao, Chaoyuan Shi, Ziqi Ye, Xiaorong Guo, Peixuan Li — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 55%

Omni LLMs can naturally locate key audio and visual evidence and integrate information across modalities for audiovisual reasoning. Existing audio-visual reasoning benchmarks that mainly rely on multiple-choice questions usually score models solely based on answer correctness, making it difficult to assess whether a model truly forms a reasoning chain from evidence to the correct answer. In addition, textual or unimodal shortcuts may allow models to answer correctly without analyzing the necessary audio-visual evidence, further weakening the reliability of evaluation benchmarks. To address these issues, we introduce EvaOmni, an evidence-grounded benchmark containing 392 videos and 1,027 purely human-annotated question-answer pairs, covering 15 capabilities and three levels: retrieval and localization, basic understanding, and complex reasoning. We design a series of rigorous annotation procedures in which experts annotate audio and visual evidence and construct video-based distractors; humans review whether questions meet strict inspection criteria, and finally models filter textual and unimodal shortcuts. Beyond high-quality question-answer pairs, we require models to simultaneously return the answer and supporting evidence intervals in order to evaluate the complete reasoning chain. We propose the Coverage Grounded Score (C.G.) to measure how much annotated evidence a model identifies, and the Bottleneck Grounded Score (B.G.) to penalize poor temporal localization in either modality; both are incorporated into answer correctness. Across 7 open-weight and 17 proprietary Omni LLMs, Gemini3.7-Flash achieved the highest overall accuracy (79.4%), C.G. (70.8%), and B.G. (35.1%). At Level 3, it reached 79.1% accuracy, but its B.G. was only 12.3%. These results show that even the most advanced models still have room for development in grounded audio-visual reasoning.

★ BlenderWorld: Benchmarking Visual Agents for 3D Creation from Real-World References

Xinyu Fang, Yuchi Wang, Weihao Yu, Rongyao Fang, Zhaohai Li, Zhixun Li — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 45%

Recent advances in multimodal large language models (MLLMs) have enabled them to act as autonomous agents, creating and editing 3D models via code or direct software manipulation. In practical workflows, these models must infer object shapes, materials, spatial layouts, and lighting from diverse references while strictly following instructions to reconstruct, modify, or design 3D content. However, existing benchmarks provide limited coverage of such open-ended requests across real-world scenarios. To bridge this gap, we introduce BlenderWorld, the first comprehensive benchmark dedicated to evaluating frontier models on realistic 3D creation tasks. BlenderWorld comprises 223 tasks spanning architecture, products, natural environments, and scientific visualization, driven by text, image, video, and structured specification inputs. To rigorously assess these complex deliverables, we establish a unified evaluation protocol: we render the submitted Blender projects from multiple viewpoints and evaluate them against 2,651 task-specific binary rubrics alongside holistic visual quality scores. Benchmarking thirteen frontier models reveals that GPT-6-Astra achieves the highest overall fused score of 79.9, followed by Opus-5 (73.0) and GPT-5.6-Sol (67.7). Ultimately, our findings reveal a critical bottleneck: while current models can parse broad instructions, they still profoundly struggle with fine-grained 3D spatial coherence and accurate structural reproduction in Blender, leaving substantial headroom for future 3D-native agentics.

★ CyberArena: Benchmarking AI Agents in Live Symmetric Attack-and-Defense Combat

Bangya Liu, Jianhong Tu, Yueyang Pan, Hanhong Zhao, Zhe Ye, Hongwei Li — Fri, 23 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 70%

AI agents are demonstrating increasingly strong offensive and defensive cybersecurity capabilities in the real world. Current benchmarks mostly evaluate agents in isolation against static targets. Real cyber operations, however, are dynamic and coupled: security postures constantly change as attackers and defenders adapt to each other in real time. Static benchmarks quickly saturate and miss an agent's ability to attack and defend simultaneously under live pressure. To address this gap, we present CyberArena, a dynamic, symmetric adversarial benchmark built around multi-round attack-and-defense combat. In isolated sandboxes, two agents compete under identical rules, acting concurrently as both attackers and defenders over continuously changing services. This converts static targets into evolving, semi-open challenges, directly testing an agent's speed and precision in discovering, exploiting, and patching vulnerabilities. To sustain the benchmark's longevity, an automated agentic pipeline continuously generates, reviews, and validates new environments. Furthermore, because pairwise contests are notoriously hard to quantify, we introduce a dynamical model that maps match dynamics into unified, multidimensional scores across offensive, defensive, and domain-specific capabilities. Evaluating 12 frontier LLM agents across 40 challenges in 466 matches, CyberArena separates model capabilities where single-agent baselines hit a ceiling. The trajectories read like a fight rather than an exam: defenders patch one hole and miss the next and certify repairs on stale evidence, while capable agents think a move ahead, planting backdoors for later rounds and, in one case, writing their own packet sniffer to watch the opponent. Beyond a benchmark, CyberArena serves as an adaptable, scalable experimental platform for future research on cybersecurity agents.

★ CiteAnchor: Source-Aware Grounding for Hallucinated Citation Detection and Resolution

Chenyi Tong, Ping Nie, Yu Zhang — Fri, 23 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Hallucination 91%, Agents 62%

Hallucinated citations make it harder to trace sources and verify the evidence behind scientific claims. Detecting them is difficult because bibliographic records can be incomplete or differ across source types and publication versions. A plausible match alone neither validates the citation nor establishes its underlying source. We introduce CiteResolve, a benchmark of 423 real-world hallucinated citations annotated for source recoverability. We further propose CiteAnchor, a source-aware grounding framework that adapts retrieval and field-level verification to each source type and separately determines citation validity and whether a unique source can be recovered. On an existing detection benchmark, CiteAnchor achieves 0.841 F1 with GPT-5.6 Luna and 0.784 with an open-weight backbone, outperforming the baseline with the same backbones by 7.9 and 36.8 points. On CiteResolve, it achieves 74.7% and 70.7% end-to-end task success, outperforming the corresponding web search baselines by 16.3 and 65.3 percentage points. Our code and data are available at https://github.com/oldwizard7/CiteANCHOR_code.

★ From Single-Actor Environments to Multi-Actor Event Worlds: Can LLMs Be Event World Models?

Juntong Ni, Zewen Liu, Bohan Wang, Lige Zhang, Wei Jin — Fri, 23 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

World models help agents assess the consequences of their actions before acting. Many real systems, however, involve multiple actors who make their own decisions, can evolve even when the agent takes no action, and are only partially observed through event logs. We formalize these systems as event worlds and ask whether large language models can predict future observations from the log and the agent's proposed actions. We build complementary benchmarks from human Diplomacy games and real Polymarket trades. Our hierarchical framework compares prediction with and without other actors' actions, then tests multi-step rollouts, predictions for actions never taken, and usefulness to agents. Fine-tuning and more data improve prediction when all actors' recorded actions are given. With only the agent's action, models struggle when other actors strongly affect the outcome, and larger models and longer histories help little. This gap also limits multi-step prediction and agent decisions, with models overestimating plan success. Yet imagined experience and initialization from world models can still improve policy learning. Anticipating other actors' actions remains a central challenge for event world modeling.

★ Large language models show systematic failures in incorporating evolving medical knowledge

Min Zeng, Zaifu Zhan, Kai Yu, Rui Zhang — Fri, 23 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Hallucination 43%

Medical knowledge evolves as clinical guidelines revise recommendations, yet whether large language models (LLMs) faithfully incorporate such changes remains unclear. We introduce ChronoEvo, comprising 1,709 recommendation-level revisions across 12 forms of medical knowledge change, to distinguish recognition of revised guidance from incorporation into clinical recommendations. In the primary model, 90.8% of responses recognized revisions and 80.2% were clinically correct, but only 58.5% fully incorporated guidance, yielding a 32.3-percentage-point recognition-incorporation gap. This dissociation persisted across 14 models and was corroborated by two clinicians' adjudicated assessment of 300 responses, yielding a 40.3-percentage-point gap in the validation sample. Changes in recommendation strength and patient eligibility were associated with retention of outdated guidance. A revision-salience intervention reduced outdated-content retention and improved clinical correctness but did not significantly improve full incorporation. These findings identify faithful incorporation of evolving medical knowledge as a distinct challenge for clinical LLM reliability, even when updated guideline evidence is available.

★ Same molecule, Different Knowledge: Cross-Identifier Binding Failures in LLMs For chemistry

Khouloud Saadi, Di Wang — Fri, 23 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, New model 87%

Chemical AI workflows routinely represent the same molecule through multiple co-referring identifiers, including trivial names, IUPAC names, and SMILES, implicitly assuming that these representations provide interchangeable access to an LLM's knowledge. To test this assumption, we first introduce a controlled benchmark and show that it does not hold. Across foundation models spanning 7B to 72B parameters, even chemistry-specialized models, co-referring molecular identifiers consistently expose different subsets of the model's knowledge. We then conduct a series of controlled behavioral, representational, and causal analyses to characterize this failure. We show that even when a model can identify a molecule from its SMILES and retrieve a fact through the corresponding name, the same fact can remain inaccessible directly from SMILES. We term this phenomenon \emph{cross-identifier binding failure}. Further analysis of internal representations, conflicting identifiers, and parameter editing shows that neither improved chemical representation competence nor modification of individual facts is sufficient to establish reliable access across identifiers. Together, these results indicate that the core failure lies in the binding between an identifier and knowledge already present in the model. Motivated by this diagnosis, we introduce Binding Self-Distillation, a simple mitigation that uses lexical identity during training to transfer the model's existing knowledge access to the structural route. This enables factual retrieval directly from SMILES at inference time, without requiring a name lookup or training the underlying model from scratch.

★ OpenAttention: One Percent Is Enough with Tail Proxies

Hyunwoo Yu, Jongbeom Lee, Hyunbin Kim, Sungjun Park, Jaemin Choi, Jincheol Yang — Fri, 23 Oct 2026 00:00:00 +0000

97% · Efficiency 97%, Evaluation 33%

Sparse attention reads only a small subset of the key-value cache, and how to treat the unread tail remains a key challenge. Discarding the tail biases the output, and sampling it as previous approaches do fails at reads of 5% and below, where the tail likely holds heavy hitters the selector missed. Our key observation is that the simplest possible proxy, one token that folds the entire tail into its count-aware mean, already yields lower output error than a sample at every read behind a practical selector. We further show in closed form that at low reads sampling nearly doubles the weight error of discarding while a single proxy stays below it. Building on these findings, we explore approximating the dense attention distribution with a few proxies that target the tail. We propose OpenAttention, a pipeline of clustering, ErrorGate and top-k selection followed by a count-aware mean proxy. To implement clustering and selection both efficiently and effectively, we design training-free Query-Geometry Clustering (QGC) and a Query-Geometry Selector (QGS), which compress the keys while respecting the query directions. Additionally, unlike prior methods that rank clusters by centroid alone, ErrorGate allocates the exact read by the estimated weight error of each cluster. On RULER-32K HARD with Llama-3.1-8B, OpenAttention retains 99% of dense accuracy at a 1% read. It also achieves state-of-the-art accuracy across models, context lengths and datasets. Our customized attention kernel, OpenAttention.flash, runs the decode attention stage 10$\times$ faster than dense attention at a 1% read.

★ Evaluating Artificial Wisdom (AW): A Trajectory-Based Framework for Context-Sensitive Decision Regulation

Md. Jalil Piran — Fri, 23 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Capability benchmarks measure what large language models can do, but not whether their recommendations remain appropriate as decision conditions change. This paper proposes a trajectory-based measurement framework and prospective validation protocol for artificial wisdom, defined behaviorally as context-sensitive regulation of judgment. Versioned reference specifications determine admissible responses and transition materiality before evaluation. Seven stage-level criteria assess judgment under current conditions, while two transition components assess material-update adequacy and commitment-preserving stability. Stability requires prior adequacy; switching within registered commitment classes is permitted, and persistent error earns no stability credit. A nine-component profile is primary, with a secondary bottleneck score and separate critical-failure reporting. Three simplified simulations examine precision, bottleneck estimation, and sensitivity to model-family count. A single-criterion simulation met its precision criteria at 24 scenarios per domain. Near a bottleneck tie, the estimated minimum had a downward bias of 0.55 points at that size. A simplified incremental-prediction simulation detected a Brier improvement of 0.027 with probabilities of .508 and .742 using eight and twelve families, respectively; it does not establish power for the operational protocol. An exploratory predecessor-rubric study comprising 71 evaluation records illustrates aggregation sensitivity but does not validate the proposed instrument. The twelve-family target remains provisional pending feasibility assessment and calibration of the complete analysis. Whether the profile predicts independent decision outcomes beyond capability remains an open empirical question.

★ Training Small Language Models for Transferable Tool Selection

Fri, 23 Oct 2026 00:00:00 +0000

97% · Agents 97%, New model 88%, Efficiency 58%

As LLM agents operate over corpora of more than ten thousand tools, selecting the right tool for a complex multi-step user instruction becomes a central bottleneck. Tool descriptions cannot all fit in context, and relevance-based retrievers only rank tools without knowing which are actually needed. Recent work instead leverages proprietary commercial LLMs like Claude or ChatGPT to drive a retriever, decomposing the instruction and selecting over retrieved candidates. We show that a small, open model can be trained via reinforcement learning to perform the same functionality. On ToolBench, a Qwen3-8B model post-trained with our recipe outperforms Llama-4-Maverick (402B). In addition, the trained model transfers directly to different retrievers and generalizes to new tool corpora in different domains.

★ Simulated Control Evals

Jeffrey Olmo, David Zhang, Tyler Tracy — Fri, 23 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 67%

Control evaluations measure whether safeguards can prevent an AI model from causing undesired outcomes even when it is intentionally trying to do so. These evaluations typically use purpose-built control settings and extrapolate their results to a model's deployment. However, the effectiveness of safeguards in deployment can depend on features of environments and tasks that control settings do not represent. To broaden where control evaluations can be run, we introduce simulated control evaluations (SCEs), which use a language model to simulate tool responses in place of an environment. We evaluate SCEs on LinuxArena, where execution provides ground-truth outcomes. We compare SCE prediction errors with those from a separate control setting, BashArena, and with variation in control metrics within LinuxArena. For independently sampled evaluation and deployment task sets, SCEs predict safety with a root mean squared error of 3.6 percentage points across five monitors at a 5% audit budget, compared with 10.1 for BashArena. SCEs closely reproduce distributions of attack strategies, attacker reasoning, and monitor reasoning, with over tenfold lower KL divergence from execution than BashArena. SCEs explain 78-85% of the between-environment variance in attack success, monitor scores, and safety. This accuracy is sample efficient: SCEs predict safety with less error than BashArena using fewer than 5% as many task-pair draws as the deployment task set. Strategies discovered through simulation also improve monitor performance on held-out LinuxArena attacks, outperforming BashArena and approximately matching improvement from executed strategies. These results suggest that deployment-specific SCEs can be more predictive than control evaluations executed in separate control settings.

★ Beyond Watching and Searching: Benchmarking Agents for Video-centered Deep Research Reports

Jiahao Meng, Qize Yang, Yuxuan Wang, Yunfei Chu, Wensi Dai, Jin Xu — Fri, 23 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 96%, Hallucination 41%

Deep research agents synthesize open-web evidence into long reports, but current benchmarks do not evaluate research centered on a source video. We introduce Video-Centered Deep Research Report Generation. Given a video and user request, an agent identifies research questions, retrieves external multimodal evidence, and produces an interactive report linked to relevant video moments. This task has direct practical value because it helps users locate relevant moments, verify claims, reproduce demonstrated procedures, and make informed decisions from video content. We present a skill-based workflow and OmniDR-Bench for this task. For evaluation, we propose Omni Atomic Rubric Trees (Omni-ARTs). Each tree contains three dimensions, 23 shared directions, and task-specific atomic binary criteria. The directions test audiovisual grounding, multimodal evidence use, and interactive presentation. Audited Rubric Construction (ARC) builds these trees through independent evidence collection, automated checks, behavioral calibration, and iterative human review. OmniDR-Bench contains 50 Chinese and English tasks across 19 realworld subcategories. It includes 6,414 criteria. Across 18 agent configurations, the best system scores 63.3 out of 100. Current agents still need broader content coverage, stronger causal and logical analysis across modalities, and more accurate citation support. We further analyze models, harnesses, skill configurations, video categories, and research trajectories. The code and data are released at https://github.com/marinero4972/OmniDR-Bench.

★ SpatialScout: Can VLM Agents Actively Explore and Map 3D Scenes?

Zhaoxi Zhang, Hao Liang, Linxuan Wu, Yiming Wu, Yifu Yuan, Kun Liang — Fri, 23 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Spatial intelligence requires agents to actively acquire visual evidence and integrate observations into a coherent representation of the three-dimensional world. Yet spatial question answering offers limited evidence of whether models can construct accurate metric maps through autonomous exploration. We introduce SpatialScout, a benchmark for evaluating vision-language model (VLM) agents on active object-level 3D mapping. SpatialScout comprises 80 diverse indoor scenes containing 3,140 target objects. Given an object inventory, RGB observations, camera calibration, and camera poses, agents control a camera to explore each scene and estimate object centers in a shared coordinate frame. The task requires coordinating viewpoint selection, cross-view association, geometric estimation, and map revision under bounded interaction budgets. Evaluation jointly measures localization accuracy and target coverage through object-scale overlap scores that penalize missing predictions. Experiments with frontier VLMs reveal substantial limitations: the strongest evaluated model, Claude Opus 5.5, achieves a mean overlap score of 41.71%, with 47.23% of targets exceeding an intersection-over-union (IoU) threshold of 0.5 and only 3.25% exceeding 0.9. These results reveal a substantial gap between recovering coarse object layouts and constructing complete, precise metric maps. SpatialScout provides a quantitative testbed for developing agents that build reliable 3D spatial representations through active perception. Code is available at https://github.com/JesseZZZZZ/SpatialScout.

★ Automating ML for Science: Can Frontier Agents Climb Scientific Hills in the Wild?

Ming Zhong, Stacy Li, Nicholas Carlini, Matthew Jagielski — Fri, 23 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 76%

Frontier coding agents are increasingly framed as end-to-end automated researchers, and ML for science offers a demanding test of that promise: improving a scientific model often requires domain insight beyond routine ML engineering. We introduce Agent4Sci, a long-horizon testbed for measuring how far today's agents can improve the published best models on scientific ML benchmarks, spanning 20 benchmarks across 11 domains. Rather than ranking agents on a leaderboard, Agent4Sci asks what kinds of progress agents can make on these tasks under in-the-wild conditions, and where they fall short. Empirically, today's frontier agents advance the published best on every benchmark, with Claude Opus reaching 10.4% mean relative improvement and no run flagged for academic misconduct. These gains, however, come almost entirely from a bag of ML-engineering tricks rather than research-level insight. Trajectory analysis surfaces three recurring limitations in this setting: novelty miscalibration, resource under-utilization, and rapid collapse into local engineering sweeps. Lightweight plug-in skills for research ideation, experiment management, and scheduled checklists raise Claude Opus from 10.4% to 20.1% and Codex from 4.4% to 8.1%, while yielding more diverse, domain-relevant interventions. Agent4Sci gives the community a way to track and push frontier agents from ML engineering toward domain insight on real-world ML-for-science problems. Code and data are available on GitHub (https://github.com/maszhongming/agent4sci) and Hugging Face (https://huggingface.co/datasets/MingZhong/agent4sci).

★ Design Anything, Design at Scale

Xuchao Ren, Siyuan Feng, Jing Huang, Yanhua Yao, Jing Wang, Lingyu Alina Wang — Fri, 23 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Hallucination 54%

As multimodal resources are transformed into media outputs, agentic AI is expected to stay aligned with design intent over long horizons. Designers follow briefs specifying goals, constraints, and deliverables, with later work building on earlier decisions. Recent work reuses design experience across tasks, yet alignment among artifacts within one project remains challenging. Agents therefore need to preserve and apply this intent across domains and as projects grow. We introduce Design Anything, a graph-based harness that preserves design intent outside the model and binds deliverables to the workflow steps that produce them through an intent-execution graph. A task contract retains constraints and decisions, selected designs become explicit inputs to later generations, and a deterministic controller uses bindings and execution records to check output types and counts, identify outstanding deliverables, and guide subsequent turns. We also introduce DesignAnything-Bench, with 228 tasks across 19 design domains, including 38 project families at scales of 6, 14, and 24 artifacts. With the LLM held fixed, Design Anything satisfies 85.0% and 92.5% of the brief's factual constraints with DeepSeek V4.1 Flash High and GPT-5.6 Sol High, respectively, exceeding the strongest same-model baselines by 15.4% and 14.4%. These gains reach 25.1% and 23.0% on 24-artifact tasks. Under each model, Design Anything leads or ties in constraint satisfaction in 18 of 19 domains. Professional designers also prefer its complete deliverable sets over those of same-model baselines in a blind evaluation.

★ Event-STVG: Spatio-Temporal Video Grounding for Long-Form Event Reasoning

Fri, 23 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Spatio-Temporal Video Grounding (STVG) aims to jointly localize a target in space and time given a natural-language query. Despite continued advances, existing STVG benchmarks remain largely centered on localizing targets through instance-specific descriptions, leaving them largely decoupled from long-form reasoning increasingly emphasized in multimodal large language models (MLLMs). In this paper, we introduce Event-STVG, a task that shifts STVG from locating what is described toward event-level reasoning by unifying long-form context, repeated-event reasoning, event localization, role inference, and multi-target grounding. To facilitate research on Event-STVG, we construct a large-scale benchmark comprising 4,016 annotated event occurrences across soccer and seven additional domains. Experiments with state-of-the-art MLLMs reveal substantial difficulty, with the strongest model achieving only 5.4% m-vIoU at the 10-minute horizon. Across domains, we consistently observe confusion among repeated same-type events, spatial grounding bottlenecks, and rare success in jointly grounding all required roles. Event-STVG establishes a new challenge for visual understanding in MLLMs, calling for future models that can jointly reason about complex events across space and time. Code and annotations are available at https://github.com/THU-KEG/EventSTVG

★ Where is Physical RSI capability from? An Observational Scaling Law of LLM-as-policy

Zekai Li, Jiayi Yu, Jinlu Zhang, Luyang Niu, Zekun Qi, Jun Zheng — Fri, 23 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

In this paper, we systematically investigate the mechanisms, scaling behavior, and generalization of physical Recursive self-improvement. We introduce PhysRSI-eval to quantify physical RSI through a five-axis metric: First-Attempt Performance, Efficiency, Success, Budget and Improvement from failure. Different VLMs are evaluated on 18 manipulation tasks under a fixed interaction budget. The analysis of experimental results motivates the Physical Barrel Hypothesis: different capabilities may constrain different parts of improvement. The five-axis metrics also distinguish systems with close public benchmark averages and show that high scores need not come with early or efficient gains. We study observational scaling from two perspectives: predicting the relative performance of VLM agents from their public capabilities, and forecasting later progress from early physical feedback. Together, these analyses ask how quickly and how far an agent can improve. Moreover, we conduct workspace interventions and then analyze what improvement leaves behind. Removing history reduces late-round success, frozen workspaces remain useful on unseen layouts, and three cross-task cases achieve success earlier. The results suggest that scene-specific adaptation and reusable experience can coexist.

★ Same Beliefs, Different Policies: Learning Reusable Representations for Early Stopping in LLM Reasoning

Fri, 23 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, New model 66%, Evaluation 49%

Reasoning LLMs often continue reasoning after reaching a correct answer, wasting computation and sometimes replacing it with an incorrect one. Deciding when to stop is a metareasoning problem: the model must weigh what further reasoning could achieve against its cost, given what it can already answer. Direct utility training learns a stopping representation at a chosen token price; supporting new prices can require additional controller training. We introduce BELT (BELief-based Termination), which separates outcome prediction from stopping control. It learns beliefs about current and future answers from a frozen LLM's hidden states, then fits a linear stopping policy on the frozen features. The prediction targets are independent of token price, so one probe supports different accuracy-cost trade-offs through CPU policy refitting or threshold adjustment. Averaged over five out-of-distribution (OOD) math benchmarks, BELT reduces generated tokens by 41% on Qwen3-1.7B and 51% on OLMo-3-7B-Think without loss of accuracy. The same math-trained BELT transfers to coding and agentic tasks without training on code or agentic data. In all three domains, its accuracy-cost trade-offs are competitive with or better than those of end-to-end controllers. On OLMo3, its probe has 21.2x fewer parameters than one checkpoint of the strongest baseline, OS-Pruner, and needs 55x fewer training GPU-hours than OS-Pruner's five price-specific checkpoints.

★ Protegrity-PII Technical Report

Fri, 23 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, New model 75%, Efficiency 61%

We present Protegrity-PII, a new PII decision model for structured and unstructured text. Protegrity-PII achieves a typed recall of 0.915 with a precision of 0.907 on a new benchmark we propose, PII-SCOPE. Our model processes text in 0.49 seconds per million characters at \$0.0005 per million characters on a single RTX 6000 using batched inference. Due in part to the expansive nature of the entity type taxonomy used in training, Protegrity-PII achieves superior out-of-distribution generalization when benchmarked against a selection of state of the art models and datasets.

★ WikiOriginQA: Automated Construction of a QA Dataset with Knowledge Origins for Analyzing Cultural Bias

Kento Haneda, Yosuke Kishinami, Ryo Fujii, Keisuke Sakaguchi, Makoto Morishita — Fri, 23 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

Existing datasets on cultural knowledge have been typically created by human annotators familiar with each culture. However, the process requires considerable human effort, presenting a significant challenge to scalability. To address this challenge, we introduce WikiOriginQA, a QA dataset automatically constructed by objectively representing the origins of knowledge through the creation dates of different Wikipedia language editions. Experimental results with the created dataset confirm that LLM performance tends to be biased toward Anglophone cultures. Furthermore, the results suggest the importance of knowledge propagation into the Anglosphere. These findings demonstrate the validity of our automated approach, enabling cultural bias evaluations with superior scalability across diverse languages without extensive manual annotation.

★ Synthetic-to-Real Code-Switched ASR: A Spontaneous Nepali-English Podcast Benchmark

Firoj Paudel — Fri, 23 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, New model 59%

Nepali-English code-switched conversational speech presents severe challenges for automatic speech recognition (ASR), chiefly through script confusion (inappropriate mixing of Devanagari and Latin orthography) and elevated error rates at language switch points. Supervised fine-tuning (SFT) can adapt multilingual foundation models, but human code-switched audio is expensive to collect and annotate. We curate a new benchmark of 13,044 spontaneous podcast utterances (15.45 h total; 10,214 test, 2,830 dev), audited and verified to eliminate machine translation leakages and acoustic loops. We investigate a two-stage synthetic-to-real adaptation framework: Stage 1 applies SFT on 0.95 h of speech synthesized via a multi-speaker, dual-engine TTS pipeline; Stage 2 explores Group Relative Policy Optimization (GRPO) with transcription and script rewards. We find that synthetic SFT provides the primary cross-domain transfer to spontaneous human speech, reducing Mixed Error Rate (MER) from 23.40% to 21.52% (-1.88 pp) and switch-point error rate (SPER) from 35.89% to 28.77% (-7.12 pp). Subsequent GRPO yields small, statistically significant adjustments on 173 test utterances (1.69%), driving English script confusion (SCR_en) from 6.45% to 6.36% (p <= 0.001) and SPER to 28.70% (p = 0.022 clustered). Crucially, ablation reveals that in this SFT-converged encoder-decoder setting, an auxiliary script reward term yields performance identical to MER-only optimization because rollout candidate groups exhibit near-zero script variance on synthetic data. On monolingual benchmarks, SFT improves Nepali MER by 2.71 pp (stabilizing at 18.42% after GRPO, a 2.38 pp gain over baseline) while preserving English lexical WER (3.38% vs. 3.33%). Verified transcripts, timestamps, episode metadata, and reconstruction scripts are publicly released.

★ ReflectPO: Reflection-Guided Policy Optimization for Long-Horizon LLM Agents

Rumei Li, Xunliang Cai — Fri, 23 Oct 2026 00:00:00 +0000

96% · Agents 96%, New model 90%, Evaluation 38%

With external tools, large language models (LLMs) can tackle complex tasks through sustained environmental interaction. LLM-based agents are therefore increasingly used to tackle complex long-horizon tasks in real-world environments. However, these tasks involve sequential, interdependent decisions and changes in environmental states. Early execution errors affect subsequent interaction histories and environmental states, and their effects accumulate along the decision chain. To address this challenge, we propose Reflection-Guided Policy Optimization (ReflectPO), a reinforcement learning framework for autonomous error recovery in long-horizon tasks. ReflectPO comprises two core components. The first is Reflection-Guided Sampling. We use branch differences and repeated visits in interaction trajectories to construct hints for error recovery. The second is Tree-Structured Credit Assignment. We organize sampled trajectories into trees and assign action-level credit using parent--child reward differences and recovery gains. Experiments over two long-horizon tool-use benchmarks show that ReflectPO outperforms multiple existing baselines in most domains. It also achieves the best overall performance on average. Further analyses show that ReflectPO effectively improves agents' ability to recognize their own deviations, recover from erroneous states, and continue making progress toward task completion.

★ IQuest-Q1: Advancing Foundation Capabilities for Agentic CLI Systems

A. Jian Yang, An Huang, Bo Dan, Chen Mu, Chengran Yang, Chuan Shi — Fri, 23 Oct 2026 00:00:00 +0000

96% · New model 96%, Agents 58%

We present IQuest-Q1, a sparse Mixture-of-Experts (MoE) agentic foundation model with 320B total parameters (15B activated). IQuest-Q1 is developed through a three-stage process: (1) stable and efficient pre-training, combining Stable SplitNorm, Muon, and muP with a two-stage curriculum whose second stage increases the coding-data mixture, preserving a wide stability margin between safe and optimal learning rates; (2) fine-grained mid-training, expanding the data mixture across domains and formats and progressively extending the context; and (3) multi-scaffold with multi-version post-training, combining SFT, domain-specialized RL, and MOPD across various scaffolds to integrate complementary training signals and strengthen agentic coding capabilities. Furthermore, model merging combines different experts and checkpoints from different stages while mitigating conflicts among their parameter updates. The resulting IQuest-Q1 demonstrates capabilities in both coding and general tasks. We release IQuest-Q1 as an open-source model to support community research and development.

★ Cola2: Co-Adaptive Hierarchical Language Modeling with Continuous Semantic Compression and Balanced Discrete Generation

Hongcan Guo, Yian Zhao, Feng Wang, Shen Nie, Guoqiang Wei, Yan Zeng — Fri, 23 Oct 2026 00:00:00 +0000

96% · New model 96%, Evaluation 35%

Hierarchical language generation separates continuous semantic modeling from discrete text realization, but its effectiveness depends on what the latent space compresses and how the discrete generator uses it. We propose Cola2, a framework that jointly studies continuous semantic compression, the balance between reconstruction and generation, and continuous-discrete co-adaptation. Cola2 dynamically segments text into semantic spans, controls the compression rate through auxiliary-loss-free negative feedback, and combines reconstruction and generation supervision to learn compact, predictive latent representations. A block-causal masked diffusion decoder is trained with clean, noisy, and dropped latent conditions, while classifier-free guidance adjusts the balance between semantic reconstruction and autonomous generation. Token-prediction gradients update the encoder through the latents, whereas Flow Matching prior learning is decoupled from representation learning by stop-gradient. During inference, generated text is re-encoded under a fixed span layout to keep the latent history consistent with the actual text context. Experiments on five downstream tasks with about 210 billion training tokens validate the core designs and show competitive, unsaturated scaling. With 346.03 billion source tokens, an unconverged checkpoint reaches a 60.10 average across eight tasks and surpasses Llama-3-8B on HellaSwag and StoryCloze despite its reported pretraining budget being over 43 times larger. These results demonstrate the potential of continuous-latent hierarchical language generation.

★ ValuSpec: Plug-and-Play Candidate Valuation before Target Verification for Tree-Based Speculative Decoding

Liang He, SiYuan Ma, Qishi Zhan, Yongqi Fan, Mingyu Cao, Qizhen Lan — Fri, 23 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, Evaluation 43%

Tree-based speculative decoding accelerates large language model inference by verifying multiple candidate branches in parallel. However, in practical serving scenarios, similar requests repeatedly induce overlapping candidate branches. While retrieval-based methods mitigate this by reusing historical drafts, the resulting candidate trees often contain many low-quality branches as the candidate tree grows, which reduces verification efficiency and wastes target-side computation. To address this, we propose ValuSpec, a plug-and-play candidate valuation mechanism that filters low-quality branches from a supplied candidate tree before expensive target-side verification. ValuSpec introduces transfer-based supervision to align filtering with the target model's behavior. A filtering threshold is estimated from the target model's greedy trajectories on a calibration split, without any parameter training. The target model then verifies the filtered tree under the output-identical greedy rule, so accepted tokens are guaranteed to match standard autoregressive decoding exactly. Experimental results on HumanEval, GSM8K, Dolly, and SpecBench show that, in our main evaluation, ValuSpec achieves up to 2.38x end-to-end speedup in retrieval-based tree settings. Furthermore, integrating ValuSpec into the EAGLE-3 dynamic tree framework yields up to 2.78x end-to-end speedup without modifying its tree construction strategy, demonstrating that ValuSpec serves as a generally applicable and plug-and-play candidate valuation module.

★ Jev-DLC: Decision-Level Continual Learning from Typed Feedback

Yangyang Feng, Jianle Gan, Zhuoyan Feng, Chengwei Qin — Fri, 23 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, New model 82%

Individual preferences often become visible only after deployment, when a user accepts one decision or revokes it in favor of another. Jev-style typed interfaces make this feedback explicit by identifying the selected action and, under revocation, the decision it replaces. We study how these consequences persist across requests as Decision-Level Continual Learning from Typed Feedback (Jev-DLC). Jev-DLC treats each correction as both evidence about a recurring action identity and a directed relation between the preferred and rejected choices. It preserves these signals in complementary bounded channels above a frozen scorer and combines their scores to shape the next decision. The method retains no raw interaction history, adds no scorer call, and performs no runtime language-model backward pass. We also introduce TCPA-Bench, a chronological suite that credits a correction only through its effect on later decisions. On TCPA-Repeat, Jev-DLC reaches 44.83% user-macro accuracy overall, 2.43 points above Recent-4 history replay. After at least two earlier feedback events, it reaches 56.07%, a 7.54-point margin over the strongest reported comparison at this depth. These results show how explicit typed corrections can become compact causal state for continual personalization.

★ How Ground Truth Shapes Retrieval Scores: Evidence-Aware Evaluation of a Retrieval-Augmented Generation System for Breast Lesions

Fri, 23 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Hallucination 55%

Purpose: To describe an evidence-aware retrieval evaluation protocol for a clinical retrieval-augmented generation (RAG) system over breast lesion management literature, addressing the dependence of retrieval scores on annotation coverage. Methods: An 18-document corpus and an 18-query development benchmark, with one Yes/No or WH-question per document, were used. Retrieval combined dense embeddings, BM25, and cross-encoder reranking. Chunk sizes of 128 to 1024 tokens, overlaps of 0 to 20%, and four retriever configurations were evaluated. Relevance was defined by evidence spans mapped to all containing chunks, treating overlapping chunks with equivalent evidence as relevant. Retrieval was evaluated using R-Precision, Recall@K, MRR, and nDCG@K. Separately, generation was assessed on a 512-token index at K=1 to 5 using DeepEval claim-level metrics, with deepseek-v4-pro as the LLM judge, and BLEU/ROUGE metrics. Results: The selected hybrid retriever, using 384-token chunks and 20% overlap, retrieved accepted evidence in the top three results for 17 of 18 queries. Recall@3, R-Precision, MRR@3, and nDCG@3 each reached 0.94. Ground-truth curation increased evidence spans from 25 to 37 and accepted chunk targets from 36 to 51 without changing retrieval rankings. Conclusion: Cross-encoder reranking achieved the highest retrieval scores. Expanded relevance annotations increased recognized evidence without changing rankings, highlighting annotation sensitivity. The small development set and retrieval-informed curation require independent validation, including radiologist assessment of evidence sufficiency.

★ Retrieval Augmented Generation Compositionality

Ido Levy, Segev Shlomov, Leshem Choshen — Fri, 23 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Hallucination 83%

Retrieval Augmented Generation (RAG) grounds large language models in external knowledge. Modern pipelines combine embedding, chunking, retrieval, reranking, query augmentation, context compression, and agentic verification, yet these techniques are typically evaluated one at a time. It remains unclear whether their isolated gains survive composition, whether retrieval quality and answer faithfulness favor the same design choices, and whether conclusions from small corpora hold at large scale. We pre-declare three hypotheses and introduce LASAGNA, an open-source framework for controlled RAG evaluation. We test 450 configurations across eight benchmarks and corpora ranging from 5K to 200K documents. Three findings emerge. First, individually strong techniques from different phases of a RAG pipeline do not necessarily compose: three retrieval techniques predict a +47% combined improvement but deliver only +17% when used together. Second, component gains stack differently for retrieval and faithfulness, and optimizing one can hurt the other. Third, the best composition changes as the corpus grows: the combination that wins on a small corpus may not remain best at large scale. These findings show that gains measured in isolation are not reliable predictors of realistic performance, and that new RAG techniques should be tested with the full surrounding stack. We call on researchers to do so and supply LASAGNA to simplify these experiments.

★ Beyond Compilation: Evaluating Faithful Natural-Language-to-Lean Statement Formalization

Fri, 23 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Hallucination 69%, Agents 48%

Lean verifies that a generated declaration is well typed, but cannot check whether it states what the user intended. We measure how far compilation overstates faithfulness in statement autoformalization without canonical Lean targets. Our criterion requires Lean compilation and agreement of two LLM judges from different model families. On an independent random audit, it agrees with the human majority on 91.5% of cases; it rarely accepts an unfaithful output but rejects some faithful ones. Across eight systems on 227 university-level statements, the gap between compiling and accepted outputs ranges from 1.3 to 29.5 percentage points: our tool-augmented agent compiles 87.2% of the statements, but only 57.7% are accepted. Human review of all the agent's rejections confirms that 23.8 of its 29.5 points are real translation errors. For the gaps of the seven one-shot systems, a two-stage review by Claude Opus 5.5 with Lean and GPT-6.1 Sol, validated on these human labels, finds that 78-100% are real translation errors for six of them. When GPT-5.2, one of the two judges, grades the translations of two models released after it, GPT-6 Astra and Claude Opus 5.5, it accepts 83% of the statements for each, against 21% for one-shot GPT-5.2. For these two models, the same review classifies only 4 of their 54 rejections (7%) as translation errors, so most of their measured gap is judge errors rather than translation errors. This is consistent with these systems erring less while the judge does not: the fewer errors a system makes, the smaller the share of its rejections that are translation errors. Finally, a 2^3 tool ablation of the agent shows that Lean feedback drives most of its compilation gain, and that nearly half of this gain fails the semantic criterion. Code and data are available at https://github.com/grenadeComing/beyond-compilation.

★ AQL-Lite: Enabling Efficient Knowledge Acquisition for Arabic LLMs via Data-Centric Continual Pretraining

Houssam Eddine Boukhalfa, Mourad Bouache, Mourad Dahmane — Fri, 23 Oct 2026 00:00:00 +0000

95% · New model 95%, Efficiency 60%, Evaluation 47%

Training Arabic large language models faces challenges from noisy web data, dialectal variation, and the high cost of continual pretraining (CPT). We introduce AQL-Lite (AQL: Arabic Quality Language), a time- and resource-efficient framework for adapting a multilingual LLM to Modern Standard Arabic (MSA). AQL-Lite couples a quality-first curation pipeline (structural cleaning, heuristic and model-based filtering, MSA selection, and deduplication) with a partial-parameter CPT recipe that trains RS-LoRA adapters across the network together with the last eight transformer blocks and the embedding matrix, under layer-wise learning rates. At a matched 1B-token budget, CPT on AQL-curated data raises a four-benchmark Arabic mean by 3.51 points over CPT on raw web text, which yields no gain over the base model. In a controlled 5B-token comparison, AQL-Lite recovers 94% of the gain of full-parameter CPT (mean 73.19 vs. 73.58) in 52% of the wall-clock time and outperforms adapter-only baselines on every benchmark, and a component ablation shows that adapters, unfrozen blocks, and embedding training each contribute. Scaled to 40B MSA tokens and instruction-tuned on one million examples, the resulting 12B model scores 75.3 on the seven-benchmark Open Arabic LLM Leaderboard v2 (OALL v2) suite. These results suggest that partial-parameter CPT on a carefully curated corpus is a practical route to efficient MSA adaptation.

★ Route-and-Generate: KL-Regularized Router Improvement for Reasoning Routing in LLMs

Fri, 23 Oct 2026 00:00:00 +0000

95% · New model 95%, Efficiency 82%, Evaluation 33%

Large language models often apply complex reasoning indiscriminately, leading to overthinking on simple tasks and underthinking on harder ones. We present Route-and-Generate (Rouge), a hierarchical training framework that enables LLMs to adaptively decide when and how much reasoning is required for each input. Our approach decouples reasoning effort allocation from answer generation by introducing a routing policy that allocates a fixed sampling budget across multiple reasoning formats, while the base model is optimized with GRPO using per-sample rewards. To promote efficient reasoning, we employ a rank-based reward that favors shorter correct solutions, and update the routing policy via a closed-form KL-regularized policy improvement step distilled into a neural router. This design encourages concise answers on easy problems while reserving longer reasoning for harder ones, achieving improved accuracy-efficiency trade-offs across multiple reasoning benchmarks.

★ What the Trace Shows and What the Label Says: Auditing Behavioral Detectors in Agent Evaluation

Idunnuoluwa A. Adeniji, Shilpika, Michael E. Papka — Fri, 23 Oct 2026 00:00:00 +0000

95% · Agents 95%, Evaluation 88%

LLM agents are often evaluated by reducing execution traces to labels such as final-answer correctness, tool verification, and fault propagation. These labels are measurements produced by parsers, behavioral detectors, and harness rules, and they may not fully capture what the trace shows. We study this in a tool-use setting built on GSM8K, where agents call a calculator whose outputs are occasionally corrupted, giving clear ground truth for what the tool returned, what the model later used, and what the detectors reported. We audit three trace-level detectors across six production models and 30 problems per model. Final correctness, visible re-verification, and fault propagation can disagree in systematic ways. A key failure mode is silent tool-output override: the calculator returns a corrupted value, but the model later uses a different, often correct, value without calling the tool again, so the bad value stops propagating while a verification detector records no re-check. Human review of 90 traces agrees perfectly with the automated labels for final correctness (67/67), fault impact (19/19), and fault relevance (18/18), but only 67/81 for verification; all 14 disagreements fall into four recurring patterns. Execution limits also change the reported result: in an ablation across two models, one ranges from 21/30 to 27/30 correct as step and token budgets change, while the other stays between 27/30 and 29/30. These findings motivate measurement transparency: an automated label should come with the trace evidence behind it, whether the question the detector asks applied to that trace, whether the run was cut off by an execution limit, and any human corrections. Our trace viewer makes this information inspectable alongside the reported results.

★ Governed Self-Evolution for Autonomous Investing: A Multi-Agent LLM System with an Invariant Core

Yebo Feng, Jingyang Dai, Liangrui Pan, Nelson Lau, Jiahua Xu, Yang Liu — Fri, 23 Oct 2026 00:00:00 +0000

95% · Agents 95%, Evaluation 54%

LLM agents that act on live Web data must combine heterogeneous evidence, remember what they have done and improve from outcomes without weakening their own safeguards. We study this problem in autonomous investing, where every action is irreversible and based on untrusted Web content. We present Steward, a multi-agent LLM system that draws the boundary of self-evolution in its type system. First, nine role-specialised agents decide jointly: four analysts read news feeds, SEC filings, macroeconomic series and market data and emit typed reports, from which a Portfolio Manager sets target positions that a protected Risk Manager audits. Second, a seven-layer shared memory, assembled by code from the system's own record, gives each decision its position context, the market narrative, calibration and the agent's past decisions. Third, the system improves itself by review: code grades every trade and forecast nightly and writes the grades into memory, and each quarter a reflector proposes prompt learnings that a deterministic gate applies reversibly and replay of recorded inputs verifies. All three operate around an invariant core of order arithmetic, risk rules and discipline-agent prompts that no self-edit can reach. In a five-month live deployment (102 trading days, 1,632 LLM calls), the system contained a 24.6% sector drawdown to a 0.3% loss; in replay of its recorded inputs, a single model given the same raw data traded more and lost 1.35 points against the market over five days where the joint design showed no significant loss, and no injected instruction was followed; and a 2015-2026 backtest shows the core alone lowering the maximum drawdown under all 50 random entry policies. The reflector traced the dominant failure to its root cause, and the gate applied a learning that the agent adopted on trigger evenings and on no control evening. Code: https://github.com/yebof/quant-agent

★ Joules-to-GDP: How Much Economic Value Do AI Systems Produce per Unit of Energy?

Fri, 23 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

An AI system can produce more economically valuable work while delivering less value for the energy it consumes. We therefore ask: How much economic value does an AI system create per unit of energy consumed? We formalize economic value in two ways: (1) net dollar value, the output's dollar value minus AI-system execution cost, and (2) time saving, the expert's completion time minus the AI system's completion time. Dividing each by measured end-to-end energy yields Net Value per Joule (NVJ) and Time Saving per Joule (TSJ), reported in USD/MJ and hours/MJ. We study 220 public GDPval tasks across 44 occupations, serving five open-weight models locally and evaluating four accelerator types. First, the occupation groups where AI systems succeed most often, deliver the greatest dollar value, or save the most time do not necessarily provide the greatest economic return per joule. For example, among the tested tasks, engineering work delivers about 10% more net dollar value per task than software development. Yet software development delivers about 27% more net value per joule. Second, GLM 5.2 delivers the highest average net dollar value, at $72.27 per task using 0.430 MJ, equivalent in energy to about four hours of active laptop use. By contrast, Gemma 4 31B delivers 56% of that net dollar value using just 23% of the energy, achieving approximately 2.5 times the NVJ and 2.8 times the TSJ. Third, across four single-accelerator deployments, Qwen3.6-27B's NVJ varies by a factor of 4.32 and TSJ by 9.27, with B200 leading both. Additionally, one and eight H100s deliver similar net dollar value, but eight reduce completion time by 22% at 3.24 times the energy. Our framework makes output value and expert time saving under an energy budget explicit objectives for improving models, accelerators, and software, while offering a new basis for assessing AI's potential economic effect on occupational work.

★ Silicon Samples Match the Answers but Not the Respondents

Yusuke Kumagae, Masahiro Kaneko — Fri, 23 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

Silicon sampling generates survey respondents by having a large language model (LLM) answer a survey as a person with given demographic attributes, and the resulting synthetic populations are increasingly used in place of human ones. Existing evaluations focus on individual-answer accuracy, per-question answer distributions, and correlations across questions. None of these compares the respondents themselves, so a generator whose respondents have combinations of answers found in no human respondent can still score well if it reproduces every answer distribution. We propose the respondent coverage gap (RCG), which represents each respondent as a point defined by their complete answer profile and measures how far each human respondent is from the nearest generated one, relative to the distance between human samples. On two US social surveys, all six LLMs we evaluate leave human respondents uncovered, and under our main settings, they fall behind a baseline that samples each answer independently from the human marginals. Across the prompting and inference settings we test, every LLM has a higher RCG than a Gaussian copula fitted to the human data.

★ DConstrain: Compiling Context-Parameterized Transitions for Diffusion Language Model Constrained Decoding

Shidong Cao, Jing Ma, Wenhao Chai, Yuchen Tian, Lun Du, Da Zheng — Fri, 23 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, Evaluation 48%, Hallucination 32%

Diffusion language models (DLMs) decode multiple tokens in parallel, so individually valid tokens can be incompatible together under a schema. Exact joint inference over a grammar graph enforces compatibility, but it requires token-labeled successors across many hypothetical parser configurations. Encoding stacks and field histories in compiled states can cause exponential growth, which forces prior exact decoders to fix property orders and exclude valid outputs. We present DConstrain, which compiles schemas into context-parameterized transition programs. Cold compilation specializes token consumption to local grammar states while retaining context-dependent stack operations, environment guards, and caller-dependent deferred parsing symbolically. Tracking field histories as runtime values supports arbitrary property order without enumerating field subsets offline. At runtime, warm compilation binds reached configurations to these programs to construct exact successors, overlapping with the model forward pass because graph construction requires no model scores. Furthermore, token-effect grouping merges tokens with identical compiled effects before runtime contexts are bound, so groups are reused across stacks and environments even when concrete successors differ. Objective-specific score aggregation and token restoration provably preserve both the Viterbi optimum and the token-labeled path distribution of ungrouped inference. Across three benchmarks and two DLM families, DConstrain compiles 100% of BFCL schemas under arbitrary property orders, whereas an optimized finite-state baseline compiles only 71.30-85.81%. Under arbitrary property order, it reduces BFCL end-to-end decoding latency by 6.65-46.26 times with 100% schema validity and raises DiffusionGemma's BFCL accuracy from 37.27% under fixed order to 81.56%.

★ RetiJev: Fast and Source-Grounded Ophthalmic Decision Models

Bo Liu, Lianyu Wang, Meng Wang, Ching-Yu Cheng, Huazhu Fu — Fri, 23 Oct 2026 00:00:00 +0000

95% · Evaluation 95%, Efficiency 95%, New model 91%

Ophthalmic clinical decision support requires fast, accurate, and privacy-preserving inference, yet autoregressive large language models (LLMs) incur substantial latency and computational overhead for decisions among predefined actions. We therefore investigate a decision-oriented alternative to generative reasoning. Specifically, we first introduce RetiJev-Data, a source-grounded ophthalmology corpus comprising 78,158 clinical-decision and diagnosis questions derived from articles in PubMed Central and a MeSH vocabulary covering 531 eye diseases, and 50,074 guideline-grading questions spanning 11 grading systems, along with a 24,662-question evaluation benchmark. Based on this, we develop RetiJev, a family of 0.8B-9B decision models that directly score candidate answers without autoregressive decoding. Across all model sizes, RetiJev outperforms both a 27B LLM and the commercial Jev 1.13 decision model on guideline grading by 10.4-23.0 points. RetiJev-9B further achieves the highest accuracy on clinical decisions, improving over both baselines by 5.4-6.5 points while matching them on diagnosis. It answers a case in 158 ms on one GPU, 3.5x faster than LLM generation and 470x faster than reasoning. Models are released at https://github.com/Awenbocc/RetiJev.

★ Harm Is Behavioural, Moderation Is Lexical: How Language and Culture Decide Whom Automated Safety Systems Protect and Whom They Silence

Gabriel Lars Sabadin, Dr Nicola Harding, Sarah Kingston — Fri, 23 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

Content warning: this paper quotes slurs and hateful statements from published test suites; the racial slur among them is masked. Automated systems decide at scale which messages are harmful, and they decide unevenly. Classifiers label African American English as offensive up to twice as often as other English, the accounts of prominent drag queens can score as more toxic than those of white nationalists, and safety behaviour that holds in English collapses in low-resource languages. We argue that over-flagging and under-protection share one cause: deciding harm from vocabulary and surface form rather than from what a message does. We organise the evidence into seven failure modes and argue, as testable hypotheses, that harmful behaviours (a request for secrecy, a demand for payment, an attempt to isolate someone) travel across languages better than the words that express them. We propose an evaluation protocol that makes the gap between the best- and worst-served language communities a headline result, and show that the naive gap is inflated under equal service and that its usual bootstrap interval is unreliable. Applied to five open systems in eleven languages, from a keyword lexicon to two current multilingual guard models, the protocol finds that none reaches native-speaker accuracy at separating harmful function from harmless form (51 of 55 system-language pairs fall short); that greater capability raises discrimination in nine languages while leaving Arabic and romanised Hindi near chance, so the gap between the best- and worst-served languages grows with capability; that one system both over-flags and under-protects the same language most; and that a language a system reads worst can look the most fairly treated when only false positives are reported.

★ Prompting Strategy Effects Depend on the Judge: A Multilingual Study of LLM-as-a-Judge Robustness

Asal Mehrabi, Mohammad Aliannejadi, Clemencia Siro — Fri, 23 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

LLM-as-a-Judge evaluation is increasingly applied to multilingual tasks, yet prompting strategies such as Chain-of-Thought and Role Conditioning are recommended based solely on English benchmarks. Whether their effectiveness transfers across judge models and languages remains unknown: prompt-sensitivity studies are largely English-centric, while existing multilingual judge studies fix the prompt template. In this work, we address this gap by measuring this interaction through a benchmark of 9 prompting strategies, 4 open-weight judges, 27 languages, and 4 tasks across 13 datasets, yielding over 1M pointwise scores on gold-reference outputs, validated against human judgments. We find that prompting strategies are not transferable: the choice of the judge model dominates the choice of prompting strategy, strategy effects reverse direction across judges, and stability degrades in low-resource languages, where these effects are concealed by aggregate reporting. These findings show that prompting effects in multilingual evaluation are conditioned on the judge, task, and language. Consequently, prompting strategies should be validated for the intended evaluation setting, and cross-study comparisons should account for both judge and prompt configuration.

★ Towards User Simulators That Stay On Task And Still Sound Natural

Mustafa Omer Gul, Philippe Laban, Tanya Goyal, Jennifer Neville — Fri, 23 Oct 2026 00:00:00 +0000

95% · New model 95%, Evaluation 68%

Users increasingly rely on AI systems for extended, multi-turn tasks, yet evidence indicates that LLM-based system performance degrades over prolonged interaction. To evaluate and improve AI assistants in such settings, we need user simulators that faithfully follow complex user intents while exhibiting realistic conversational behavior. Existing approaches manifest a trade-off: prompted assistants can follow complex user intents but lack behavioral fidelity, while specialized user simulators better approximate conversational behavior but struggle to remain faithful to complex intents. In this paper, we propose a reinforcement learning (RL) recipe that jointly rewards task intent adherence and behavioral naturalness. A crucial contribution of the method is operationalizing naturalness with a lightweight discriminator trained in an alternating, adversarial loop with the user simulator. We release the resulting model, UserLM-2.0, which produces valid simulated conversations (i.e., meeting both task-faithfulness and naturalness criteria) 46.2% of the time, over three times the rate of the strongest prior user simulator (14.1%).

★ BeNice! Closing Looping's Inference-Memory Gap via KV-Sharing

Hayden Prairie, Jacob M. Ryan, Taylor Berg-Kirkpatrick, Daniel Y. Fu — Fri, 23 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, New model 90%

Looped architectures scale quality while keeping parameter count fixed by repeatedly passing activations through the same block of layers. During inference, each recurrence generates independent KV-caches, causing memory to grow linearly. To address this challenge, prior work has proposed KV-sharing (i.e., sharing KV-caches across depth), finding that, in some cases, the KV-cache of the final recurrence can be shared with the earlier recurrences without quality loss, a property we denote as KV-shareability. However, the mechanisms underlying KV-shareability remain poorly understood, with even identically trained looped architectures not consistently exhibiting it. In this paper, we study how KV-shareability can be enabled in looped models. We propose a two-stage method: in the first stage, we use stochastic depth during pretraining, which induces contractive residual dynamics that drive representations toward a common limit across recurrences. In the second stage, we propose BeNice!, a post-training method that broadcasts the final recurrence's KV-cache into the first recurrence, training it to produce compatible residuals while preserving the original model's quality. As a result, BeNice! enables KV-shareability while matching standard finetuning's unshared validation perplexity. Using BeNice!, we show that KV-sharing is complementary to alternative KV-compression methods, composing with memory-efficient sequence mixers (e.g., multi-latent attention, Kimi delta attention) and low-precision KV-cache quantization (e.g., fp8, mxfp8, nvfp4). Finally, in simulated low-memory, long-context inference, KV-sharing improves cache hit rates by up to 31.4%, increases throughput by up to 23.6%, and reduces p90 time-to-first-token by up to 83.6%.

★ Specter: Fast MoE Inference via Quantization-Enabled Multi-Step Speculative Offloading

Wenfeng Wang, Xiaofeng Hou, Xinfeng Xia, Xinkai WAng, Mingxuan Zhang, Peng Tang — Fri, 23 Oct 2026 00:00:00 +0000

95% · Efficiency 95%

Mixture-of-experts (MoE) models often exceed a single accelerator's memory capacity. Offloading experts to host memory creates a severe PCIe I/O bottleneck because data-dependent routing keeps transfers on the critical path. We present Specter, an inference system that uses quantization to combine speculative decoding with expert offloading. Its lightweight quantized draft model runs efficiently on device while preserving high-fidelity expert activation patterns. This provides accurate multi-step lookahead over future expert usage, allowing the runtime to organize expert movement and execution ahead of verification and improve both communication and execution efficiency. Under single-request inference, Specter achieves an average 1.50x (up to 1.88x) speedup in time per output token (TPOT) over the state-of-the-art system. Specter makes memory-constrained, single-GPU MoE inference more practical on commodity hardware. Our code is available at https://github.com/wwwwwf/specter_asplos.

MCP vs CLI for AI Agents: A Comparative Analysis of Tool Integration Strategies for Autonomous Systems

Fri, 23 Oct 2026 00:00:00 +0000

94% · Agents 94%, Evaluation 42%

AI agents require interfaces to interact with external systems such as files, APIs, databases, and cloud services. Two competing paradigms have emerged: CLI (Command Line Interface), where agents execute shell commands directly, and MCP (Model Context Protocol), a structured JSON-RPC protocol providing governed, typed tool access. This paper presents a comparative analysis across 14 evaluation dimensions, supported by worked examples, security assessment against established compliance frameworks, and economic modeling. Our analysis indicates that CLI provides superior efficiency for single-user, local-only workflows, while MCP addresses requirements that CLI fundamentally cannot satisfy, including fine-grained access control, credential isolation, structured audit, and multi-tenant governance. We further examine how progressive disclosure techniques eliminate MCP's historical weakness (context window overhead), yielding token efficiency comparable to CLI while preserving its architectural benefits. A hybrid architecture is recommended for production systems. This analysis draws upon academic, government, and industry sources including NSA formal guidance (CSI U/OO/6030316-26), OWASP security standards, and economic modeling from BCG and Gartner.

Self-Assessment Reinforcement Learning for Search-State Awareness in Retrieval-Augmented Large Language Models

Taewhoo Lee, Minju Song, Jungwoo Park, Chanwoong Yoon, Junha Jung, Hyeon Hwang — Fri, 23 Oct 2026 00:00:00 +0000

94% · Agents 94%, New model 92%, Hallucination 83%

Recent advances in search agents have enabled large language models to tackle complex tasks by retrieving and synthesizing external information. However, existing search agents often lack search-state awareness, i.e., the ability to recognize what information has been obtained and what remains missing, leading to hallucinated reasoning and prematurely terminated search. To address this limitation, we propose Self-Assessment Reinforcement Learning (SARL), a reinforcement learning framework in which models learn to recognize their own search state during training and improve their search behavior accordingly, without requiring additional assessment at test time. SARL achieves this through a dual reward design that captures the corrective effect of each assessment while penalizing excessive reliance on it, jointly encouraging effective and efficient search trajectories. Under this setup, SARL achieves a 24.3% relative improvement over the strongest baseline, with average gains of 35.9% on multi-hop reasoning and 27.9% on deep research benchmarks against the strongest per-benchmark baselines. SARL further surpasses methods built on auxiliary supervision or multi-agent coordination, despite operating as a standalone RL framework with no inference-time scaffolding. Beyond final performance, SARL maintains training stability under extended search horizons, a regime where existing methods consistently degrade, enabling sustained gains as task complexity grows.

ULME: Predictive Coupling for Mutually Reinforcing Embedding and Language Modeling

Yang Liu, Zhiqiang Yuan, Shu-Dong Huang, Jiancheng Lv, Liwei Yang, Qiwei Yan — Fri, 23 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 33%

Text embedding and text generation are two of the most useful capabilities of modern large language models, yet they are typically provided by separate specialized models (an embedder and a generator) trained with incompatible objectives and deployed side by side in retrieval-augmented generation and agent systems. Recent unified embedding-generation models such as GritLM share parameters between the two capabilities, but they either switch attention regimes between modes or optimize embedding and language modeling as parallel objectives that never supervise each other. We argue that these designs are only partially unified: the two capabilities coexist on the same weights, but do not shape each other. We propose ULME, a post-training framework that unifies embedding and language modeling on a causal decoder-only LLM through predictive coupling. ULME inserts a chain of learnable slots between the query and the answer, reads a semantic embedding and a reasoning embedding from two designated slots, and trains three objectives on a shared causal backbone: a contrastive loss anchoring the semantic slot to document representations, a language-modeling loss supervising answer generation, and a stop-gradient predictive loss that aligns the reasoning slot with a pooled summary of the answer, so that generation actively shapes what the embedding encodes. Across two model families (Qwen2.5, Qwen3) and eight scales spanning 0.5B-8B parameters, a single ULME checkpoint improves MTEB by 3.3-4.1 points over the strongest open-source baseline, tightens BRIGHT and AdvBench-IR simultaneously, and remains on par with a generation-only fine-tune on standard accuracy benchmarks (MMLU, GSM8K, MMLU-Pro, HumanEval), demonstrating that genuine unification yields mutually reinforcing rather than merely coexisting capabilities.

GIFT: Gradient Informed Filtering for Distillation Augmented Reasoning RL

Shijie Li, Yilin Gao, Tieyuan Chen, Yun Luo, Weiyao Lin, Hang Yu — Fri, 23 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 32%

Training large language models for complex reasoning tasks relies on two complementary signals: sparse, outcome-grounded reinforcement learning (RL) for correctness, and dense, token-level distillation for guidance. However, complementarity in supervision does not guarantee compatibility in optimization; imitation-focused distillation can conflict with the outcome-driven RL objective, and existing hybrids lack a principled resolution. We propose \textbf{GIFT} (\textbf{G}radient-\textbf{I}nformed \textbf{F}il\textbf{T}ering for Distillation-Augmented Reasoning RL), a novel method that directly addresses the allocation problem of which teacher updates to apply. GIFT performs a token-level geometric analysis, decomposing the distillation gradient into a component \textbf{parallel} to the RL gradient, which measures first-order conflict and provides a \textbf{safety} filter, and a component \textbf{orthogonal} to it, which quantifies novel directional guidance and provides a measure of \textbf{value}. This decomposition yields a principled scoring rule to rank token-level guidance. We combine this with a progressive \textbf{coverage annealing} schedule that shrinks the budget of tokens receiving distillation, ensuring that the most conflicting updates are discarded first while the most complementary ones are retained at full strength. This facilitates a smooth transition from dense, geometrically-filtered guidance to a pure, outcome-grounded objective at convergence. Extensive experiments on mathematical and logical reasoning benchmarks show that GIFT consistently outperforms strong baselines across multiple model scales, achieving relative gains of up to 9.5\% on difficult mathematical reasoning tasks. Code: https://github.com/lilifree/GIFT

Diagnosing the Explicit-to-Applied Theory of Mind Gap in LLMs

Prajvi Saxena, Sabine Janzen, Wolfgang Maass — Fri, 23 Oct 2026 00:00:00 +0000

94% · Evaluation 94%, New model 88%

As large language models are deployed in high-stakes roles such as therapy, education, and personal assistance, their social cognition must be assessed reliably. Current Theory of Mind (ToM) benchmarks report aggregate scores that conflate two distinct competencies: explicit ToM (recognizing what someone believes) and applied ToM (using that belief to predict behavior). Building on this distinction, we investigate the gap and its recoverability using four structurally diverse benchmarks. Aggregate reporting obscures uneven competence: on SimpleToM, Llama-3.1-8B scores 51.5 percent overall while recognizing mental states at 99.4 percent and predicting behavior at only 28.9 percent. To assess recovery, we compare structured prompting, matched in-context learning, and our proposed Belief-State Test-Time Training (BSTTT), which adapts lightweight parameters on answered support questions through action reconstruction. Across Qwen2.5-7B, Llama-3.1-8B, and GPT-oss-20B, recovery depends on the model, task, and intervention. On SimpleToM, reminders combined with BSTTT improve Llama's behavior prediction by 40.8 percentage points; reminders account for most of the overall improvement, with BSTTT adding 1.3 to 5.5 points beyond reminders across models. Contextual examples often outperform weight updates, while BSTTT remains stronger in specific model-benchmark settings. Temporal belief tracking remains difficult despite partial recovery. These findings show that aggregate ToM scores can overstate applied competence while hiding performance recoverable through targeted assistance. We recommend decomposed reporting of explicit and applied accuracy alongside intervention comparisons that establish which failures are recoverable and under what conditions.

Certified Execution Harness for Agentic AI: Safe and Least-Intrusive Petri-Net Supervisory Control Between Reasoning and Action

Chao Gu, Quanyan Zhu — Fri, 23 Oct 2026 00:00:00 +0000

94% · Agents 94%

This paper studies the problem of safe tool execution for large language model (LLM) agents under prompt injection. Using Petri nets, we formalize the execution harness between the LLM and its tools as a supervisory control problem, in which marks represent capabilities, approvals and budgets. Since a transition fires only if the marks it needs are present, an injected request whose preconditions are unmet or already spent is refused regardless of its wording. Treating tool outcomes as uncontrollable, we enforce safety by the conventional maximally permissive supervisor, and this safety transfers to the running system under four assumptions. Moreover, to capture an agent that may propose again after a refusal, we introduce the notion of a replanning closed loop, in which each retry costs an inference call. We show that, in this loop, the conventional supervisor can be arbitrarily more intrusive than a stricter safe one. Furthermore, for a given proposer model, the maximally permissive least-intrusive supervisor is characterized as a value-threshold refinement of the conventional one. Under precise conditions, this refinement orders two concurrent actions by at most one monitor place. On an autonomous release agent, we find by synthesis that safe deployment requires a step that the supervisor can refuse after a rollout starts.

SQuAT: Spherical Lattice Vector Quantization-Aware Training

Fri, 23 Oct 2026 00:00:00 +0000

94% · New model 94%, Efficiency 53%

Vector quantization (VQ), in particular lattice-based methods, offer strong rate-distortion trade-offs for low-bit large language models (LLMs), yet their behavior under quantization-aware training (VQ-QAT) remains poorly understood. We scale lattice VQ-QAT to billion-parameter LLMs and multi-billion-token training regimes, and introduce SQuAT, a shape-gain VQ-QAT approach trained with a separate straight-through estimator (STE) for the shape and gain. SQuAT decomposes weights into direction and magnitude: normalized directions are quantized using an $E_8$-derived spherical codebook, while gains are quantized independently with low-bit scalar quantizers. We show that, unlike a standard STE applied to the full quantizer, which ignores the shape-gain structure, these separate STEs induce tangent-space updates for directions, yielding an implicit form of spherical (Riemannian) optimization. To enable scalability, we develop a distributed exact nearest-neighbor assignment scheme, in which workers quantize disjoint weight blocks and communicate only discrete indices, reducing VQ-QAT training time by 12.8$\times$ and allowing us to scale it to 10B tokens. Across instruction-tuned LLaMA-3.2-1B/3B, Qwen3-4B, and base Qwen3-1.7B, SQuAT consistently outperforms standard VQ-QAT STE under identical quantizers and initialization, and achieves SOTA VQ-QAT accuracy.

When to Follow the Teacher: Advantage-Based On-Policy Distillation

Mahdi Sabbaghi, George Pappas, Hamed Hassani, Adel Javanmard — Fri, 23 Oct 2026 00:00:00 +0000

94% · New model 94%

Online Policy Distillation (OPD) often outperforms offline methods such as Supervised Fine-Tuning (SFT) by training on trajectories sampled directly from the student. However, standard OPD trains the student to follow the teacher at every state of the trajectory, assuming that teacher guidance is always beneficial. We instead distinguish between what the teacher prefers and what actually helps the current student succeed. We introduce Advantage-Based On-Policy Distillation (AOPD), which weights teacher actions by their counterfactual advantage under the student. Teacher actions that improve the student's expected outcome are reinforced, while those that do not are downweighted. Directly computing these advantages requires trying alternative teacher actions at each step and running new rollouts from each. Instead, we show that the AOPD objective can be estimated using only standard student rollouts and their terminal rewards. Our estimator uses teacher importance weights on the student-sampled actions to provide unbiased advantage-weighted supervision, without any additional rollouts or reward evaluations. This allows AOPD to focus teacher supervision on actions that help the current student while preserving the rollout efficiency of standard OPD. Experiments on math, coding, and Medqa show gains over distillation and RL baselines in both external-teacher and self-distillation settings.

Edit, Don't Imitate: Minimal-Edit On-Policy Distillation for Multimodal Reasoning

Wen Ye, Ailin Deng, Zirui Wang, Shixing Yu, Han Zhang, Tian Yun — Fri, 23 Oct 2026 00:00:00 +0000

93% · New model 93%, Hallucination 41%, Efficiency 34%

Visual mathematical reasoning can fail before reasoning begins: when a vision-language model (VLM) misreads a diagram, subsequent reasoning proceeds from false premises. This perceptual bottleneck is especially pronounced in smaller models, while accurate textual descriptions can substantially narrow the gap to larger models. Because obtaining such descriptions at inference requires additional annotation or model calls, we propose minimal-edit corrective on-policy distillation. Using reference figure descriptions and verified reference solutions as training-only information, a teacher scores incorrect student rollouts under a minimal-edit objective, and only teacher-disfavored tokens contribute to the corrective loss. Correct rollouts receive task-reward reinforcement learning. This preserves on-policy supervision while avoiding full-response imitation. Across model scales, the method requires no descriptions at test time and meets or exceeds the corresponding base model supplied with a reference figure description at inference. It also improves perception-oriented performance, particularly low-level geometric perception, and transfers to external mathematics and perception benchmarks. Because the minimal-edit teacher evaluates corrections rather than generating its own solution, the method remains effective with a frozen same-size teacher, although its gains are smaller than with a larger teacher on most metrics.

Better Reasoning, Not Just Bigger Models: A Structure-Grounded Framework for Efficient and Reliable Inference

Amirhossein Aghaei, Liangyan Li, Shiyi Yang, Shaohua Wang, Yan Wang — Fri, 23 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 81%, Hallucination 50%

Scaling large language models has substantially improved linguistic fluency, yet reliable autonomous agents require more than fluent generation. In multi-turn inference, current agentic systems often produce semantically plausible but structurally ungrounded reasoning, especially under incomplete observations. We propose Structure-Grounded Uncertainty Reduction (SGUR), a training-free framework that complements token-level generation with active uncertainty reduction over structured hypothesis spaces. SGUR represents incomplete observations as a query graph and aligns them with candidate hypothesis graphs through a prior-weighted partial graph matching mechanism, where database-derived empirical priors assign greater importance to decision-relevant relations rather than treating all graph edges uniformly. The resulting inference state is iteratively refined by an active information-gain strategy that selects the next most informative node to query, systematically pruning the candidate space. We instantiate SGUR on a challenging PrimeKG-derived sequential medical diagnosis benchmark, demonstrating that structural grounding can partially substitute for model scale: starting from only two observed symptoms, SGUR achieves 92.5% Hit@3, outperforming the strongest Graph-RAG baseline by 9.4 percentage points while using approximately 5x fewer tokens. Notably, SGUR enables smaller backbone models to match or exceed the performance of models up to 10x larger, suggesting that structural grounding provides a more robust and efficient paradigm for scaling agent performance than relying exclusively on increasing model size.

Compiling Autoregressive LLM Control into Policy-Exact Parallel Inference for Hierarchical RAG

Zehao Lin, Qianfeng Shen, Xixuan Hao, Shaobo Cui, Chunyu Li, Kai Chen — Fri, 23 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 54%

Large language models increasingly act as controllers in structured decision pipelines, yet their decisions are still executed by autoregressive text decoding. Hierarchical RAG is one such setting: at each frontier of a document tree, an LLM commits to $b$ of $k$ sections under an input-dependent action grammar, executed as either failure-prone free-form JSON or reliable but strictly serial grammar-constrained generation. We show that this serialization is unnecessary for bounded-horizon control: the reachable future forms a shallow action-prefix tree, which we compile --- one root prefill plus one tree-causal probe evaluate every prefix-conditioned action distribution with serial decoding's tokens, positions, and causal ancestry. A prefix-state lemma makes compiled control policy-exact in exact arithmetic: the same policy with a two-forward critical path and zero output tokens. One further probe pass certifies the exact order-marginalized MAP set. Across four document-disjoint splits and two scales on the primary MetaX device, compiled control matches the serial policy on $84$--$89\%$ of questions --- every instrumented divergence a floating-point near-tie --- and cuts control latency by $35$--$47\%$ versus token-wise serial decoding and $6$--$7\%$ versus an idealized chunked engine; an H800 repetition preserves the ordering.

Disentangling Robustness and Sycophancy in Multilingual LLMs

Rhitabrat Pokharel, Ameeta Agrawal — Fri, 23 Oct 2026 00:00:00 +0000

93% · Evaluation 93%

When a multilingual LLM agrees with a user's false premise, is it being sycophantic, or does it simply not know better in that language? Current research does not distinguish these cases: they record only whether the model agreed, not why. We introduce the SC-matrix, that crosses a model's competence (whether it answers correctly under no pressure) with its sycophancy (whether it maintains that answer under pressure). Our results show that (1) identical aggregate sycophancy rates can mask different failure compositions, (2) high competence in a language does not always protect against sycophancy, and (3) reasoning-oriented mitigations outperform directive-based strategies. We find that multilingual sycophancy is not a single phenomenon but a composite of different failures that require distinct diagnosis.

UniGuard: Multi-View Guardrail Learning for Cross-Interface Safety Generalization

Fri, 23 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 80%

Guardrail models are increasingly used as safety layers for LLMs and agentic systems, but most are trained on fixed prompt or response interfaces and become unreliable when the same safety case appears as a conversation, agent trajectory, multi-agent trace, or policy-conditioned input. We propose UniGuard, a training framework that learns interface-invariant safety representations from multiple semantically aligned views of the same safety case. We also construct the UniGuard-FT dataset by transforming 1,000 public safety cases into prompt, response, multi-turn conversation, agent-trace, and multi-agent-trace views, and train UniGuard-0.6B with classification, supervised contrastive learning, and asymmetric confidence calibration losses. Across unseen-interface benchmarks ATBench, PolicyGuardBench, and DynaBench, UniGuard-0.6B improves over its pretrained guardrail by 45.12 F1 points on average and reduces expected calibration error by 35.41 points, while remaining competitive with much larger guardrails such as Claude-Sonnet-4.5 and GPT-5.2 - despite being a 0.6B model and trained only with synthetic data.

TESSA: A Governance-Oriented Custom GPT for SME Guidance under the EU AI Act: Design, Implementation, and Evaluation

Fri, 23 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, Hallucination 48%

The increasing adoption of generative AI presents new opportunities while raising AI governance and regulatory challenges, particularly for small and medium-sized enterprises (SMEs) navigating the European Union Artificial Intelligence Act (EU AI Act). To support AI governance education, this study developed TESSA, a governance-oriented educational artifact within the OpenAI Custom GPT environment. Following a design science research approach, TESSA was evaluated against baseline ChatGPT using 100 benchmark scenarios through governance alignment assessment, human expert review, and consistency testing. Preliminary LLM-assisted assessment showed governance alignment rates of 90% for TESSA and 83% for baseline ChatGPT, while independent validation by an AI regulation law practitioner yielded 71% and 70%, respectively. Human expert review indicated that both systems provided generally strong governance guidance, while TESSA maintained high consistency across repeated executions. The findings suggest that governance-oriented grounding can support structured, practical, and stable AI-assisted compliance guidance for SMEs. The study contributes (i) a governance-oriented Custom GPT artifact for AI governance education, (ii) a human-validated regulatory update mechanism, and (iii) a multidimensional evaluation framework for governance-oriented AI artifact.

MOER: Mix-Policy Optimization under Expert Prefix Guidance with Selective Rollback

Xuyang Zhong, Jiuding Yang, Yiwen Guo, Chen Liu — Fri, 23 Oct 2026 00:00:00 +0000

93% · New model 93%

Mix-policy optimization incorporates off-policy expert data into the reinforcement learning (RL) loop to transcend the capability boundary of on-policy RL. However, existing methods either provide static full-trace guidance that suppresses self-exploration, or adopt expert prefix guidance without correctness guarantees, leading to suboptimal performance or even training collapse. In this paper, we propose MOER, a mix-policy optimization framework that harmonizes expert guidance with self-exploration. MOER injects a decaying expert chain-of-thought prefix into the prompt and lets the model complete the remaining reasoning, with the prefix length progressively annealed to encourage self-exploration. A selective rollback mechanism replaces failed completions with the original full expert trace, guaranteeing the correctness of every off-policy trajectory. Furthermore, a harmonized mix-policy objective optimizes on-policy tokens with a GRPO loss while supervising both expert prefix and completion tokens with a trust-region SFT loss, substantially stabilizing training. Extensive experiments show that MOER achieves the highest average scores in mathematical and general domain reasoning tasks, outperforming all basic and mix-policy baselines, and generalizes consistently across diverse tasks, model scales, and base model categories. Codes are available at https://github.com/CityU-MLO/MOER.

A Reality Check of Inference-Time Context Augmentation Methods

Haeun Yu, Geert Heyman, Pepa Atanasova, Isabelle Augenstein — Fri, 23 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, Hallucination 68%

Large language models (LLMs) often fail to utilise the context given in the input, resulting in hallucinations or a fallback to parametric knowledge. Context augmentation methods aim to improve model performance by modifying the input, from inserting filler tokens that add computational time to adding relevant content such as elaboration of a context during inference. However, these methods have been evaluated in isolation and only by their performance gains, leaving it unclear how they compare and whether they reliably improve performance. In this paper, we present an evaluation framework spanning five methods that differ in how informative the inserted content is and apply them at the sentence level across seven models spanning multiple sizes on three datasets. We find that accuracy gains are not always genuine improvements but undirected perturbations: the methods correct some predictions while breaking others, particularly when the inserted text allows only for more computational time. Our analysis reveals that methods that bring no additional content primarily change model predictions near the decision boundary. Our findings highlight the need to evaluate the reliability of the accuracy gains before adopting them in practice.

Dual Exam: Controlled Evaluation of Multimodal Agents and World Models

Zhefan Rao, Meng Chu, Xuan Zhang, Jingfan Chen, Guohao Cai, Runyi Li — Fri, 23 Oct 2026 00:00:00 +0000

93% · Evaluation 93%

World Models (WMs) can support Multimodal Agents (MAs) through simulated experience, while Agent trajectories can inform World Model refinement. Evaluating this interaction requires separate evidence about action selection and state evolution. We introduce Dual Exam, which tests Agents under fixed environment response mechanisms and World Models under fixed action interventions across physical, digital, social, and scientific domains. The Agent track compares 10 configurations on 1,088 items in 15 reporting categories. The World Model track contains 400 cases across eight scenarios; 15 configurations have full coverage and one covers six scenarios. GPT 6 Astra leads the case-weighted Agent Index at 31.3, and GPT-Image-2 leads the full-coverage WM Overall at 21.7. These scores use different tasks and metrics and are not directly comparable. A 150-case pilot of one coupled pairing records zero task successes and motivates component-level diagnosis. Results reveal uneven task performance and failures to preserve task-relevant state. Their value for predicting coupled success or sustained reciprocal learning remains to be established.

nnInteractive Agent: General-Purpose AI Agents as 3D Image Annotators

Fri, 23 Oct 2026 00:00:00 +0000

93% · Agents 93%, Evaluation 86%

Interactive 3D segmentation models such as nnInteractive turn a few spatial prompts (points, scribbles, boxes) into accurate volumetric masks, but a human still does the annotation work: finding each structure, placing prompts, checking the result and correcting it. General-purpose AI agents can now interpret images, call tools and write code. We show that such agents can take over this work without additional training: the user makes a request in plain language, and the agent works out what has to be segmented, locates the targets, prompts nnInteractive, inspects the 3D result, corrects it and responds to feedback. Asked for 'the muscles used for chewing', an agent identifies and segments the masseters and pterygoids; asked to count liver lesions, it writes its own code to find candidates and splits a merged lesion on request. Our open-source nninteractive-agent provides nnInteractive as tools over the Model Context Protocol (MCP), together with a skill that encodes how experienced users prompt it. On a benchmark spanning CT, MRI, PET and micro-CT of humans, a mouse and an ant (14 images; 51 targets from organs and tumours to muscles, foot bones, vessel trees and an insect brain), Claude Code and Codex reach a mean Dice of 0.81 and 0.84 from the request alone; when told where each target is, both reach 0.85 and prompt nearly as effectively as an oracle that places prompts using the reference segmentation. A text-prompted 3D model (VoxTell) is more accurate on most standard radiological targets but less robust: it returns an empty mask for one of them and fails on the leg muscles, the foot bones, the mouse and the ant, where the agents remain accurate. Code: https://github.com/MIC-DKFZ/nninteractive-agent

RIFT: KV Cache Compression for Draft Models with Complementary Compressibility Training

Fri, 23 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, New model 79%

Speculative decoding is a popular decode acceleration method. As such, it is particularly relevant to long-output generation, which requires an increasingly large KV cache. KV cache compression offers a way to constrain costs in long-context drafting. However, compression methods that focus on prefill information can lose relevance as the output length grows. We term this phenomenon decode staleness and show that it degrades draft model accuracy under selection-based KV cache compression. We introduce RIFT, a family of selection-based KV cache compression strategies that balance prefill information and recent generated tokens. We show that refreshing the KV cache selection during token generation significantly improves the drafter's quality at the cost of additional decode latency. To alleviate the refresh costs, we propose an adaptive KV refresh strategy that reduces the number of score calculations while retaining the drafter's quality. We further demonstrate how a compressibility-driven drafter training phase, based on sparse attention masks, can improve the drafter's quality under a compressed KV cache. Combined with compressibility training, RIFT enables Qwen3.5-4B, drafting for Qwen3.5-122B, to retain over 98% of its full-cache acceptance length on 32K-token inputs, 4K-token outputs, and an 8K-token KV cache budget.

Post-Audit Mirage: Identifiability Limits of Offline Agent Verification

Aman Behera, Payal Payal — Fri, 23 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, Agents 42%

Offline tests can fail to reveal whether an agent update will remain safe after deployment. We operationalize this post-audit mirage using matched safe and harmful lifecycles that expose identical verifier-visible evidence while an isolated scorer records opposite deployment truth. Our theorem applies standard observational-equivalence and indistinguishability ideas to this executable audit boundary and states the resulting error-abstention tradeoff. The Post-Audit Mirage benchmark makes that boundary executable across authorization, scheduling, and shared-capacity environments. IdentifiedRangeMonitor uses fresh group-labelled harm outcomes to return deploy, hold, cannot determine, or unsupported. Experiments show that monitoring can recover the missing distinction in favorable settings but may remain undecided for rare groups, small effects, or missing observations. Verifier-agent evaluation must therefore test both calculation validity and whether the available evidence supports the deployment decision.

Prompt Sensitivity in Tool-Augmented Model Comparisons

Mateusz Modrzejewski, Tomasz Radzikowski — Fri, 23 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, Agents 87%

Language models are compared on question-answering benchmarks that pose each question in one wording, and a reworded question can change which of two models scores higher. We ask whether giving the models tools makes such comparisons more stable. An accuracy gain cannot answer this, because tools change several steps between the question and the scored answer at once. We propose a matched protocol that separates those steps: the answer readout, reasoning, the decision to call, the tool's answer, and the report supplied without a call. We apply it to six open models in five domains under five wordings. We find that tools raise accuracy where a program answers the question and widen the wording sensitivity of many pairwise comparisons. The widening follows how often a model calls, and call frequency follows the wording for some models and not others. It also depends on the interface and the instructions: a native tool-calling interface and a second tool description leave no detected widening in molecules, and sensitivity persists where tools return intermediate results. These findings show that deterministic reports do not guarantee stable comparisons. We recommend reporting call frequency and comparison stability under several wordings beside accuracy.

Trion-1: Unifying Language, Protein Sequence, and Structure with Diffusion Language Models

Tianyu Lu, Kaicheng Yang, Jiangtao Feng, Keyue Qiu, Yawen Ouyang, Yixin Wu — Fri, 23 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 39%

We envision a general-purpose protein foundation model as a unified, interactive model of sequence, structure, and function, capable of understanding, generation, and cross-modal translation while retaining the instruction-following and reasoning capabilities of modern LLMs. Toward this vision, we introduce Trion-1, the first native, unified, multimodal diffusion language model (dLLM) for proteins. Trion-1 is built on two key design choices: (1) a unified discrete token space that jointly represents sequence, structure, and natural language; (2) a dLLM backbone that replaces strict left-to-right factorization with bidirectional or semi-bidirectional token interactions, enabling global residue communication and long-range contact modeling that are naturally suited to structure modeling. This formulation unifies six core capabilities: folding, inverse folding, sequence understanding, structure understanding, sequence design, and structure design, while retaining the native language capabilities of general-purpose LLMs. On established benchmarks, Trion-1 achieves broadly competitive performance. In folding, Trion-1 achieves a TM-score above 0.8 with a single sample under strict homology control, matching or exceeding the existing tri-modal model using 200 samples. For sequence understanding, Trion-1 performs on par with top-tier LLMs such as Claude Opus 4.7 (62.37%). These results support well-aligned sequence-structure-language representations and strong cross-modal generative capabilities.

RIPPLE: Recovery Internalization through Process-Level Policy Learning and Effect-Aware Credit for Long-Horizon Agentic Coding

Jingxuan Xu, Zihang Xu, Yuechen Luo, Ziqi Miao, Wenqiang Zhu, Weihao Li — Fri, 23 Oct 2026 00:00:00 +0000

93% · New model 93%, Agents 53%, Evaluation 52%

Long-horizon coding agents can observe a failure and generate a plausible critique yet continue along the same incorrect trajectory. We call this the Critique-to-Action Handoff Gap. We introduce RIPPLE, a recovery-oriented training framework that internalizes process-level guidance into a single self-recovering policy. A prefix-only teacher provides evidence-grounded recovery supervision without access to future actions or terminal outcomes. After lightweight distillation, reinforcement learning assigns stage-specific credit to reflection triggering, diagnosis, and corrective action. Local validity signals gate trajectory advantages so that intermediate recovery decisions do not indiscriminately inherit terminal outcomes. Across four repository-level software-engineering benchmarks, RIPPLE improves the base policy by an arithmetic mean of 4.8 score points, with no auxiliary critic at deployment. Controlled ablations isolate the benefits of local, stage-aligned credit, and matched-prefix interventions show that critique semantics affect subsequent correction. Recovery gains concentrate where the prefix contains sufficient diagnostic evidence and the repository state remains reversible within the remaining budget.

An Empirical Study of Recent Large Language Models for Knowledge Graph to Text Generation on Standard Benchmarks

Priyanka, Renaud Delbru, Souhail Bakkali, Micka\"el Coustaty, Antoine Doucet — Fri, 23 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, Hallucination 82%

Knowledge graph-to-text (KG-to-text) generation verbalizes a set of knowledge graph triples into fluent natural language. While fine-tuned encoder-decoder models have long been the standard approach, recent LLMs are becoming increasingly competitive. We benchmark recent LLMs against fine-tuned encoder-decoder models, using reference-based metrics and graph-grounded reference-free LLM-based evaluation validated against human annotations. We observed that recent LLMs substantially narrow, and in several settings close, the gap with fine-tuned models on reference-based metrics, while performing better on graph-grounded metrics. LLM-based evaluation scores show limited correlation with reference-based metrics, suggesting that the two approaches capture different aspects of generation quality. Graph-grounded evaluation reveals that prompted models produce fewer contradictions and triple omissions than fine-tuned models, and these errors increase more rapidly with graph size for fine-tuned models, highlighting differences between the two approaches that reference-based metrics may not fully capture.

What Makes a Good Unmasking Policy? A Controlled Study of Learned Unmasking in Diffusion Language Models

Jinya Sakurai, Xun Xu, Jaehong Yoon — Fri, 23 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, New model 81%, Evaluation 48%

Masked diffusion language models generate tokens in parallel, but their efficiency depends on which predictions are committed. Learned unmasking policies improve the speed--accuracy trade-off, yet differ in supervision, inputs, and capacity, obscuring which choices matter. We view supervised policies as distilling a decoding procedure through a teacher and examine independently what the teacher labels as committable, on which states the policy imitates it, what the policy observes, and how much the number of function evaluations (NFE) must decrease to offset policy overhead. We find that labeling predictions matching the final output is overly permissive at low NFE, and hidden states are weak alone but can complement confidence. About a 10% NFE reduction suffices to offset policy overhead for policies up to 7.5M parameters across tested settings. We therefore train a policy on trajectory-contraction targets, which accept only commitments staying on a reference trajectory, using confidence and hidden states. On LLaDA-8B-Instruct, the resulting 856K-parameter policy achieves the best accuracy--NFE frontier and highest throughput on GSM8K, is competitive at moderate targets on MATH-500 and HumanEval, and largely transfers to Dream-7B-Instruct.

AgentVQA: A Unified Benchmark for Agentic Visual Understanding

Advait Gupta, Sachin Konan, Tom Yin, Gabriel Sarch, Zhuang Liu — Fri, 23 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, Agents 43%

Vision-language models (VLMs) can perform a broad range of tasks across diverse settings, yet their performance in agentic contexts remains poorly understood. Although online evaluations in simulators and real-world environments are essential for measuring interactive agent performance, they are costly and challenging to scale. Complementary offline VQA benchmarks can enable faster and more reliable comparisons across agentic domains, yet existing benchmarks remain limited and fragmented. To address this gap, we introduce AgentVQA, an offline benchmark for evaluating VLMs on realistic agentic scenarios across five agent domains: Web Agents, Robotics, Egocentric Videos, Games, and Spatial Planning. We build AgentVQA by manually selecting 1000 challenging, real-world agent examples from 23 source datasets and adapting them for reliable offline evaluation. The resulting benchmark preserves realistic scenarios from interactive environments across diverse domains, and its model rankings track those from corresponding online evaluations. Across the 22 evaluated models, the best achieves only 58.5% accuracy. Our cross-domain analysis shows that errors are driven primarily by failures to ground and maintain accurate representations of the current and evolving visual state, including localizing relevant UI elements, interpreting motion, tracking state changes over time, and inferring 3D spatial relationships, rather than by task misinterpretation or downstream logical reasoning.

DiffResearch: Accelerating Deep Research with Diffusion Language Models

Alan-Barsag Gazzaev, Alexey Gavrilov, Mikhail Mozikov, Sergey Muravyov — Fri, 23 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Agents 91%, Evaluation 58%

Deep research systems iteratively plan, gather evidence, synthesize, and revise, making end-to-end latency a key bottleneck for interactive use. Almost all production systems are built on autoregressive (AR) language models and inherit left-to-right decoding in planning and long-form writing. We present DiffResearch, an open multi-agent deep research framework that places a diffusion language model (dLLM) at the core of every stage, and measure how a diffusion backbone changes latency and report quality. Holding the agentic scaffold, retrieval code, prompts, and evaluation fixed, we swap the backbone between an open diffusion model and a comparable AR model. Across 100 Deep Research Bench tasks, the diffusion backbone is roughly 6 times faster end-to-end, driven by order-of-magnitude reductions in planning and writing. We further introduce dFast, a decomposition mode that reads intermediate denoising states and extracts subqueries before generation finishes, providing a 1.99 times median speedup for decomposition in isolation with no loss of report quality. On the benchmark's RACE rubric, dFast obtains the best overall score among the three configurations. Finally, in a public system-level leaderboard snapshot, an instance built on a stronger diffusion model ranks above several commercial systems. We release DiffResearch under an Apache-2.0 license.

HERA: Harness-Environment Co-Evolution for Reliable Agentic Abstention

Han Luo, Bingbing Wen, Guang Yang, Zora Zhiruo Wang, Pan Lu, Lucy Lu Wang — Fri, 23 Oct 2026 00:00:00 +0000

93% · Agents 93%, Evaluation 70%, Efficiency 68%

Large language model (LLM) agents are increasingly capable of acting in complex tool-use environments, yet they often fail to recognize when tasks are infeasible and no valid solution exists. Recent work has formalized this reliability gap as the problem of agentic abstention, and existing approaches typically optimize a model or agent harness against a fixed set of tasks, leading to limited generalization to unseen failure modes. We introduce HERA, a framework for harness-environment co-evolution for agentic abstention. HERA consists of (i) a pipeline to automatically construct verifiable pairs of feasible and infeasible tasks by applying controlled environment mutations that transform solvable tasks into cases requiring abstention, and (ii) a co-evolution procedure in which performance failures on previous tasks are used to drive harness adaptation and generate new execution environments and tasks geared towards previous weaknesses. On held-out evaluation tasks, an evolved harness from HERA improves abstention accuracy from 61.7% to 83.3% while improving feasible-task completion from 68.3% to 76.7%, achieving the highest abstention and feasible-task completion among the compared methods. The resulting best harness transfers across 19 other LLMs, improving abstention accuracy by 15.3 percentage points on average without any model-specific optimization, and enabling smaller models to match the performance of more powerful models at an estimated 85% lower cost.

MoEZip: Routing-Aware KV Cache Compression for Sparse Mixture-of-Experts LLMs

Minsang Kim, Seungjun Baek — Fri, 23 Oct 2026 00:00:00 +0000

92% · Efficiency 92%, Evaluation 62%

Mixture-of-Experts (MoE) language models scale model capacity efficiently by activating a small subset of parameters per token. However, in long-context inference, memory remains a major bottleneck because the KV cache grows linearly with sequence length. Existing KV cache compression methods reduce this cost for dense transformers, but overlook a problem specific to MoE: KV eviction can change which experts are selected during answer generation, degrading the generation quality. We propose MoEZip, a routing-aware KV cache compression for sparse MoE language models. To estimate the sensitivity of expert routing to perturbation, we propose a novel metric based on the Fisher information matrix of routing distributions. MoEZip combines this metric with attention weights for routing-aware scoring of contexts. Finally, we compose a complementary scoring system that combines eviction methods for dense and sparse-MoE architectures. Experiments show that MoEZip achieves robust performance under aggressive cache compression on various benchmarks. At a KV retention ratio of 10\%, MoEZip improves over the SoTA baseline by +24 and +20 points on average on LongBench and RULER, corresponding to 2.3$\times$ and 4.9$\times$ higher scores, respectively.

Spectral Smoothing and Spiking Perturbations Reconfigure Capability Profiles and Selective Task Performance in Large Language Models

Qince Luan, Jianhui Xu, Yang Xu — Fri, 23 Oct 2026 00:00:00 +0000

92% · Hallucination 92%, Evaluation 69%

Singular-value spectra encode important structure in transformer feed-forward projections, yet their causal role in downstream behavior remains underexplored. We study post-hoc spectral smoothing and spiking interventions on the up_proj and down_proj matrices. Positive perturbation stretches deviations from the geometric mean, making large singular values larger and small ones smaller (spiking), and vice versa. The intervention preserves singular vectors and the geometric mean of the singular values. Across five open Llama and Qwen instruction models, we evaluate the perturbed systems on capability-labeled subtasks, standard and chain-of-thought reasoning benchmarks, judge-scored truthfulness, hallucination, and safety tasks, and TRAIT personality probes. To interpret score shifts, we fit responses in a three-dimensional capability basis consisting of factual knowledge, language understanding, and deductive reasoning. The induced changes suggest structured rather than purely random sensitivity: large positive perturbations more reliably suppress language and deductive responses, moderate negative perturbations can sometimes improve factual or task-local performance on the evaluated grids, and sensitivity varies strongly by model family, benchmark module, and underlying layerwise spectral inequality. These results position spectral perturbation as a lightweight intervention for probing how capabilities are organized and traded off in large language models.

Governance-State Corruption in Long-Term Memory for LLM Agents: A Cross-Model Stress Test

Bowen Liu, Youting Wang, Bonan Shen, Dingyan Shang — Fri, 23 Oct 2026 00:00:00 +0000

92% · Evaluation 92%, Hallucination 36%

An LLM agent can retrieve a memory correctly yet use it incorrectly after its validity state changes. We study governance-state corruption: faults that alter which retained records enter model-visible context while stored content and tasks remain fixed. Across six providers, 731 benchmark-grounded bases yield 2,080 matched fault pairs. In offline control-plane checks, corruption changes the operative projection in 96.25% of pairs, while oracle-backed enforcement restores it in 100%. For downstream evaluation, a deterministically selected, non-probability LongMemEval/Memora panel resolves 298 of 300 matched triplets (238 distinct bases). Clean-minus-corrupted contrasts are 40.54 percentage points in LongMemEval binary correctness (95% base-cluster interval [30.32, 50.71]) and 30.60 points in rescaled Memora FAMA ([23.31, 37.69]); both sources also show recovery. OAKS Novel native option parity holds for all 6,714 resolved outputs. Full-cohort frozen-scorer corruption/recovery effects (-27.55/+27.51 points) remain descriptive proxies, not validated native semantic population effects. These bounded findings motivate evaluating memory control-state integrity alongside retrieval accuracy.

One Teacher, Many Views: Reducing Reference Sensitivity in On-Policy Self-Distillation

Khouloud Saadi, Di Wang — Fri, 23 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 37%

On-policy self-distillation (OPSD) improves reasoning by training a model on its own trajectories with dense token-level supervision from a privileged copy conditioned on a reference solution. Existing methods typically use a single teacher distribution conditioned on the complete reference. We identify a limitation of this formulation: a reference is only one realization of the reasoning process. Although it provides evidence about the answer, it also encodes an ordering, phrasing, and solution path. Full-reference conditioning can therefore produce correct targets that depend excessively on information unavailable to the student at inference. We show that replacing one valid reference with another, while holding the problem and student state fixed, can substantially alter the teacher distribution, a phenomenon we call reference-induced target instability. To address this instability, we introduce Monte Carlo Privileged Distillation (MCPD), which models privileged context as a distribution over partial reference views and distills the marginal teacher distribution. MCPD evaluates the teacher under multiple views and averages the resulting next-token distributions. We show that this probability-space average is the exact target induced by minimizing the expected teacher-to-student KL divergence over privileged views. Across mathematical and scientific reasoning benchmarks, we show that MCPD produces more reference-consistent supervision and consistently improves over full-reference OPSD.

Conservative Meta-Agent Search: Architecture Adaptation under Non-Stationarity

Ali Atiah Alzahrani — Fri, 23 Oct 2026 00:00:00 +0000

92% · Evaluation 92%, Efficiency 60%

Automated agentic-system design is typically evaluated on static task suites with immediate scalar rewards. Deployed systems face persistent shift, delayed dependent feedback, correlated component failures, and costly architecture changes. We formalize meta-agent architecture search as sequential incumbent replacement with switching costs and hard risk constraints and introduce Conservative Meta-Agent Search (CMAS). CMAS retrieves archived architectures and proposes typed graph edits, then scores each against the incumbent by paired shadow replay under common random numbers, with interventional component credit. Promotion requires an empirically calibrated, variance-penalized simultaneous score above a pre-specified margin, and is followed by a simulated canary stage and pre-specified rollback. Since backtests cannot certify meta-level skill, we introduce MetaInvest-Bench, a regime-switching simulator whose specified worker competence makes a surrogate-optimal architecture computable, with leakage-controlled replay, adversarial governance tracks, and five negative controls. Financial streams serve as a risk-sensitive testbed for meta-level generalization; we claim no alpha. Across 96 streams, CMAS cuts normalized mean surrogate dynamic-oracle regret from 0.116 to 0.104 and raises held-out utility from 0.44 to 0.48 over the strongest adaptive baseline; when extra workers add no value, it contracts to near-single-agent teams. Its deployment inference cost is ~6% lower, though its search-inclusive token cost is the highest of the methods compared. The observed per-update false-promotion rate is 3.4% against a nominal 5% target, and 4.8% in a sealed rerun. The calibration is empirical: it concerns an update-level replay statistic on an author-designed benchmark, and guarantees neither full-utility improvement, nor absence of simulated canary-stage harm, nor sequence-level error control.

TraceForge: Distilling with Your Own Thinking

Yan Wang — Fri, 23 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 66%

Using an LLM to generate training data for another is among the most general ways to train one, and it remains common practice. Some frontier teams have closed that route for their own models: what comes back is an answer, or a short summary, without the detailed reasoning trace behind it. That is a little ironic: they were themselves built on distilling human knowledge, and distilling an open-source model is still entirely routine; the restriction applies to their own outputs alone. Recent work recovers the hidden reasoning by replaying encrypted blocks across models. It somehow works, but it breaks the providers' terms of service, so it is not something a product can be built on long-term. We built the alternative. We take the context and the teacher's answer as input, and generate the reasoning trace ourselves, in the style of the student's own model family. Training is two-stage, supervised and then adversarial: a generator learns to regenerate the family's own reasoning traces from their context and answer, and a critic learns how real a reasoning trace looks. Applied to a teacher's answer, the generator writes one in the family's style. An independent judge rates how likely traces are to be genuine on a scale from 0 to 1: ours average 0.68, real traces 0.66, and traces selected by the critic from four candidates 0.82. These are authenticity scores, not correctness rates. We test the idea on five benchmarks with everything else held fixed. On general tasks the five training corpora are about the same; on coding, the generated reasoning beats Fable 5's own thinking content. With the same 5,000 questions and the same teacher answers, HumanEval+ improves by 3.7 points and MBPP+ by 7.1 at the final checkpoint, and the MBPP+ lead holds at every checkpoint. We suspect that reasoning written in the student's family style is easier for it to learn from.

FlowState: Execution State as Memory for Long-Horizon LLM Agents

Minghao Li, Bangyan Li, Zifan Wang, Yulong Li, Hu Xu, Gan Zhang — Fri, 23 Oct 2026 00:00:00 +0000

92% · Efficiency 92%, Agents 86%, Evaluation 35%

Long-horizon tasks require LLM agents to continually draw on information from earlier interactions. However, retaining the full history increases context costs, while compressing it risks losing details needed later, and the relevance of historical information often becomes apparent as the task progresses. To address these challenges, we propose FlowState, which treats execution state as memory that can be retained and revisited across requests, unifying current decision-making with the reuse of historical information. FlowState preserves semantically typed state nodes, their relations, and references to raw tool observations, separating persistent retention from on-demand access. Within a single execution loop, Incremental State Update (ISU) maintains the current state based on new inputs and feedback, while Progressive State Access (PSA) progressively reveals historical states and supporting evidence as needed during reasoning. Together, these mechanisms enable agents to reassess prior decisions in light of new information and guide subsequent actions. Compared with a full-context baseline using the same DeepSeek-V4-Flash model, FlowState improves the average success rate on MemoryArena and the average pass rate on τ^3-Bench by 4.55 and 13.95 percentage points, respectively, while reducing total token consumption by 43.2% and 40.6%. These results demonstrate the performance and efficiency advantages of FlowState on long-horizon tasks.

Training the Whole Agent, Not Just the Model

Lingxiao Tang, Wei Tao, Yiwen Guo, Lingfeng Bao — Fri, 23 Oct 2026 00:00:00 +0000

92% · New model 92%, Agents 91%

Training coding agents typically updates the policy while keeping the harness fixed. During training, trajectories reveal flawed repair strategies and interaction problems that vary across tasks and change as the policy learns. We introduce HarnessAdapt-RL, which trains the policy and adapts its harness during reinforcement learning. For flawed repair strategies, a frozen teacher generates task-specific mechanism-level skills. For interaction problems recurring across tasks, a separate teacher model evolves tools and hooks, keeping each change only if it improves the observed pass rate or the observed trainable rate without lowering either. The policy trains with the adapted harness, and its new trajectories guide further adaptation. With a 9B base model, HarnessAdapt-RL improves pass@1 over GRPO by 9.8, 12.6, and 6.4 percentage points on SWE-bench Verified, Multilingual, and Pro, and outperforms GRPO with the human-designed OpenHands harness. Gains from mechanism-level skills persist after their removal at evaluation. Re-evolving tools and hooks as the policy changes outperforms freezing early versions. Removing them at deployment, in contrast, costs up to 7.7 percentage points. Training therefore produces a model-harness pair: the trained policy and the evolved tools and hooks, with skills removed at deployment.

Holi-World: Building Holistic World Intelligence from Human-Ego Experience

Hao Li, Xiyuan Zheng, Yuanyuan Gao, Yalun Dai, Xinhao Ji, Xuan-Li Chen — Fri, 23 Oct 2026 00:00:00 +0000

92% · Evaluation 92%, New model 76%

Human egocentric experience offers a rich foundation for learning how the world is organized and how it changes through interaction. Realizing this potential at scale requires turning growing volumes of recordings into learning resources that connect scene structure, object semantics, and interaction histories across viewpoints and time, while reducing dependence on costly manual annotation. We introduce HOLI-WORLD, an integrated ecosystem connecting experience capture, world representation, automated data construction, model learning, and independent evaluation. To sup- port learning at scale, HOLIWORLD-Foundry combines complementary static and dynamic perception to automatically construct world records that connect persistent scene structure and evolving object states with interaction histories and supporting observations. HOLIWORLD-Data organizes these records from approximately 100 self-collected scenes into grounded learning resources, including approximately 80K QA pairs used to train HOLIWORLD-VL-4B. Independent human annotations of heldout recordings form HOLIWORLD-Bench, supporting evaluation of both annotation quality and learned spatial and temporal understanding. Experiments on ScanNet, ScanNet++, Aria Digital Twin, and HOLIWORLD-Bench demonstrate strong performance in static perception and dynamic tracking, while training on HOLIWORLD-Data substantially improves HOLIWORLD-VL-4B's spatial and temporal reasoning on HOLI- WORLD-Bench and EgoDynamic4D. HOLIWORLD provides a practical foundation for expanding learning resources from the continuous flow of human experience and advancing machine understanding of the dynamic world.

State World Models: Execute the Rules, Generate the Pixels

Chan Yan Ki, Yanran Zhang, Wenzhao Zheng, Xiaofeng Wang, Zheng Zhu, Lei Chen — Fri, 23 Oct 2026 00:00:00 +0000

92% · Evaluation 92%

Existing world models can generate action-conditioned video from a single image. Many focus on camera movement, allowing users to explore the generated scene. However, navigation alone does not ensure that an action follows persistent rules when its outcome depends on state that is invisible in the current frame. We present Stateful Harness, a framework that separates world evolution from video rendering to support stateful and rule-governed interaction. It maintains an evolving world state under specified rules, renders each resulting event with a frozen video model, and uses the generated clip to update the world state for the next turn. At each turn, the harness resolves the action selected by the user and compiles the resulting event into an instruction, while a preview clip from an editable 3D scene provides spatial and camera guidance. To evaluate rule compliance, we build WorldState-Bench, a benchmark of 50 counterfactual twin pairs that share an initial frame and action but differ in hidden state and required outcome. Under rule-only prompting, the best of the 11 evaluated video backends produces the required endings for both cases in 32% of pairs. For MiniMax-H3, specifying the required endings raises this rate from 32% to 64%, while the full harness reaches 92% with the same frozen model. These results support separating rule execution from video rendering for world interactions with hidden state.

Building Presentation Intelligence for High-Quality Slide Generation

Yuhang Xie, Cheng Li, Chenliang Li, Mingyu Xu, Mingwei Sun, Suhang Zheng — Fri, 23 Oct 2026 00:00:00 +0000

91% · New model 91%, Evaluation 85%

PowerPoint presentations (PPTs) are a dominant medium for communication across diverse domains, yet high-quality PPT generation still depends heavily on labor-intensive human effort. Recent advances in large language models (LLMs) for agentic tasks have created both new opportunities and challenges (Chen et al., 2026) for automatic PPT generation. In this work, we present a unified training framework for automatic PPT generation toward presentation intelligence, integrating cross-validated data synthesis, guided teacher enhancement, and verifiable reward optimization. We first construct a cross-validated pipeline for data synthesis, with explicit control over both difficulty and validity. We then develop a teacher-enhancement framework grounded in rules and prior knowledge to substantially strengthen the instructional capacity of teacher models. Building on this, we further improve supervised fine-tuning (SFT) through a mixed-teacher strategy and a lightweight outcome-driven trajectory filtering method. Finally, based on the resulting SFT models, we introduce a suite of verifiable aesthetic metrics together with a GRPO-based reinforcement learning framework to optimize slide-level visual attributes beyond text-only supervision. Experiments show that our enhanced SFT pipeline improves Qwen-3.6-27B and Qwen-3.6-35B-A3B by 7.45 and 9.35 points on PresentBench, respectively, surpassing GLM-5.2 while maintaining strong generalization. Reinforcement learning further improves performance across five aesthetic dimensions, compensating for the limitations of SFT in visual refinement. Overall, the results demonstrate the effectiveness of our framework for high-quality PPT generation.

Interpretive Variation and Discourse Structure in Detecting Enthymemes in Controversial Political Discourse

Martial Pastor, Nelleke Oostdijk — Fri, 23 Oct 2026 00:00:00 +0000

91% · Evaluation 91%

Enthymemes are arguments that leave a premise or conclusion unstated, and they are a common persuasion device on social media. Because readers must fill in the missing part themselves, they often disagree on whether a text contains one. This paper asks how to detect and evaluate enthymemes when such disagreement is part of the task. On tweets about vaccination and immigration, each labeled by five annotators, F1 stalls around 0.67 for every fine-tuned model we train, regardless of data size, added features or label type. We therefore ask how these models should be evaluated instead, and propose metrics that compare predictions with the full distribution of annotator votes, together with a model that predicts this distribution directly and captures disagreement better at no cost in F1. We then ask what role discourse structure plays. Models learn to rely on explicit argumentative structure: they reliably recognize tweets without an enthymeme but rarely commit on tweets that contain one. Both the models and a discourse parser miss arguments whose links are left implicit and must be inferred by the reader. Finally, we discuss what these blind spots mean for content moderation systems trained on majority labels.

Learning What to Ask for Tomorrow: Active Preference Acquisition for Long-Term Personalized Web Agents

Yiqi Wang, Zizhi Zeng, Jiaqi Zhang, Yiqun Duan, Mingkai Zheng, Taotao Cai — Fri, 23 Oct 2026 00:00:00 +0000

91% · Evaluation 91%, Agents 90%

Web agents increasingly serve the same user across repeated online tasks such as booking travel, shopping, and scheduling, so their long-term usefulness depends on how well they know the user's preferences. Passive memory and personalization methods cannot obtain preferences the user never reveals, and clarification methods that value a question by its benefit to the current task overlook the benefit of reusing the answer in later tasks. We study Active Preference Acquisition (APA): which missing preferences are worth the cost of asking? We define the Prospective Personalization Value (PPV) of a missing preference as its expected cumulative utility gain over future tasks minus the effort and disclosure-risk cost of asking, and select questions under a per-interaction budget. We characterize when cumulative reuse justifies acquisition despite insufficient single-task benefit, and evaluate this regime with an initial-profile diagnostic. APA computes the counterfactual gain of each candidate preference on public task prototypes with the agent's own task solver and aggregates it by the expected reuse of each task family. We evaluate it on ActiveWebPref, a controlled longitudinal benchmark with hidden but queryable user profiles and exact task utilities. At a horizon of 20 tasks, 34.4% of budget-feasible valuable missing preferences are amortization-dependent; APA acquires 80.3% of them, whereas current-task value of information acquires 5.1%. Under per-user asking-cost caps matched to APA's realized expenditure, APA achieves higher net utility than four heuristic selection baselines. In our cost setting, asking about every unknown preference yields lower net utility than not asking, while APA recovers 84.2% of its utility gain with 27.6% of its questions. The value of acquiring a user preference should account for its reuse across future tasks, not only its benefit to the task at hand.

From Structured Records to Patient Narratives: Preserving Clinical Alignment in LLM-Generated Health Summaries

Bayram Ayadi, Florian Geissler, Johanna Schmidhuber — Fri, 23 Oct 2026 00:00:00 +0000

91% · Hallucination 91%, Evaluation 79%

As wearable and digital health systems increasingly generate longitudinal physiological records, communicating these observations in patient-accessible language remains a significant challenge. Large Language Models (LLMs) offer a promising mechanism for translating structured clinical information into readable narratives, but this process may introduce semantic drift, where clinically precise meaning is distorted during simplification. We investigate evidence-grounded health summarization for menstrual health monitoring using a constrained generation framework that combines structured evidence conditioning with Retrieval-Augmented Generation (RAG) using a prompt-based constrained generation framework. We evaluate semantic alignment using Natural Language Inference (NLI), atomic factuality metrics, and readability measures. Across three LLMs and 4,500 generated summaries, explicit grounding constraints consistently improve clinical alignment, particularly by improving uncertainty preservation and, once faithfulness is evaluated against retrieved evidence, generally preserving or improving faithfulness, while effects on omission remain model-dependent. We also find that improved clinical alignment comes at the cost of increased reading difficulty and medical terminology. These findings highlight the need to jointly evaluate faithfulness, uncertainty preservation, and readability when developing patient-facing medical LLM systems.

Same Luck, Different Choice: Twin-World Credit Assignment for Agentic Reinforcement Learning

Zeropy Surio, Tobias Engel, Mei-Lin Chou, Anika Raman, Samuel Okafor — Fri, 23 Oct 2026 00:00:00 +0000

91% · New model 91%, Efficiency 74%, Evaluation 62%

Agentic reinforcement learning rewards a complete episode, yet the policy needs credit for every decision inside it. Counterfactual branching estimates this credit by re-running the episode after an alternative decision, but the branch draws its sampling noise independently of the factual episode, so the outcome difference mixes the effect of the decision with the luck of everything that follows. We observe that in agentic search the retriever, the corpus, and the grader are deterministic, so an episode is a deterministic function of the policy's decisions and its sampling noise. Twin worlds are therefore realizable: the same noise, a different decision. We introduce TWICE (TWIn-world Credit Estimation), which keys the Gumbel noise of every token by turn and line, rolls out counterfactual twins that inherit the factual episode's downstream noise, and turns their paired outcome differences into turn-level advantages for policy-gradient updates on actor-sampled trajectories; only the actor is deployed. Twin credit is unbiased for any coupling quality, its variance is that of an independent counterfactual scaled by $1-\rho$, where $\rho$ is the twins' outcome correlation, and under coupling the number of rollouts needed to resolve an advantage of size $\Delta$ falls from $\Theta(\Delta^{-2})$ to $\Theta(\Delta^{-1})$. A square-root rule spends the twin budget where outcomes are predicted to disagree per unit cost. On seven question-answering benchmarks, TWICE improves the average exact match of GRPO by 5.3 and 4.5 points on Qwen2.5-3B and 7B, outperforms tree-search and counterfactual-branching baselines, and uses 68% fewer auxiliary tokens than VinePPO.

Can an Agent Debug a Web Corpus? Agentic Auditing of Slovak FineWeb2

Marek \v{S}uppa, Ivan Vykopal, Jaroslav Kop\v{c}an — Fri, 23 Oct 2026 00:00:00 +0000

91% · Agents 91%, Evaluation 62%, Efficiency 49%

Multilingual pipelines such as FineWeb2 apply one recipe across a thousand languages, leaving the failure modes of individual languages unanticipated. We use an LLM once, as a corpus debugger rather than a classifier: a tool-using agent explores the corpus through a read-only, budgeted API and compiles each finding into an executable test, which a native speaker then validates. On the 26.5M-document Slovak portion of FineWeb2, three runs costing \$2.21-\$2.84 each found seven findings over four phenomena. The largest is a machine-generated content farm of 1,408,437 documents (5.32%) that passed the pipeline's filters, 86.6% of it passing even stricter thresholds. Blinded labels support 6 of the 7 findings (defect rates .53-1.00 against .08 in a random control) and do not support one. The compiled suite scans the full corpus in 0.7 s of wall-clock time without a GPU, against \$12k-\$90k per LLM classification pass. The best static sampling baseline validates more selectors than any agent run, at under half the cost. The agent contributes exact prevalence, a precision estimate checked against sampled documents, an account of why the pipeline missed each slice, and a reusable test. We release the framework, Slovak test suite, validation data and agent trajectories.

PACE: Periodic and Channel-Aligned Encoding of Four-Bit Recurrent States

Patrick Haller, Filipe Laitenberger, Alan Akbik — Fri, 23 Oct 2026 00:00:00 +0000

91% · Efficiency 91%

Linear-attention language models keep their context in a recurrent state of fixed size, where an inference system stores one copy of this state for every sequence it decodes at the same time. Storing the state at lower precision therefore lets more sequences share an accelerator. Existing state quantizers round the state after every generated token and report that four bits lose accuracy unless the bit width is raised or the model is fine-tuned.\par We introduce PACE, a post-hoc four-bit state quantizer. PACE rounds the state only once every few generated tokens, as chunked kernels already do once per chunk during prefill, and keeps the tokens generated in between in a small buffer. Rounding less often reduces the error that accumulates in the state. Additionally, we find that the large entries of the gated-delta state gather in columns, because some channels of the update written at each token remain large across many tokens. PACE therefore shares each quantization scale within one column. It also keeps the heads that retain rounding error longest at eight bits. On four hybrid models from 4B to 35B parameters, PACE with mixed four- and eight-bit heads is competitive with eight-bit DAMP on the three larger models at about two thirds of its memory, PACE with four-bit heads throughout scores higher than Q-Mamba at equal memory, and rounding every few tokens preserves the decoding speedup of the buffer.

Harness Scaling for Generalizable Agentic Reinforcement Learning

Fei Bai, Huatong Song, Chuan Hao, Ran Tao, Jian Yang, Bryan Dai — Fri, 23 Oct 2026 00:00:00 +0000

91% · Agents 91%, New model 82%, Evaluation 51%

Agent harnesses have become a central component of modern agentic systems, shaping how models use tools, manage context, and coordinate long-horizon interaction. Realizing the benefits of these harnesses depends on how effectively the underlying model can make use of them, motivating reinforcement learning directly through the harness. However, existing approaches typically train through a single fixed harness, which can lead to specialization and limited transfer across harness configurations. To address this, we introduce Harness Scaling, a new scaling dimension for agentic RL that extends training from a single harness to diverse execution configurations. Concretely, we construct a factorized harness space, select a compact training subset with balanced factorial coverage, and organize cross-harness training to combine global coverage with controlled variation for the same task. Across multiple benchmarks, model scales, and RL algorithms, Harness Scaling consistently improves generalization to unseen white-box configurations. Moreover, models trained through structured white-box variation transfer effectively to unseen external black-box harnesses, outperforming models trained through a single complex black-box harness. These results establish harness diversity as a meaningful scaling dimension for agentic RL and highlight the importance of how this diversity is organized for cross-harness generalization.

Looped Uniform Diffusion Language Models

Saiteja Utpala, Chieh-Hsin Lai, Yuki Mitsufuji, Praneeth Vepakomma — Fri, 23 Oct 2026 00:00:00 +0000

91% · Efficiency 91%, New model 88%

Uniform diffusion language models (UDLMs) generate text by iteratively denoising sequences of ordinary vocabulary tokens, any of which may be corrupted and subsequently revised. During training, UDLMs must denoise from a context that may itself be unreliable, whereas masked diffusion language models (MDLMs) always condition on a clean context. This challenging training objective makes UDLMs more FLOP-intensive to train than MDLMs, but it yields stronger downstream reasoning performance. Recently, selective looping, which recursively applies a few early Transformer layers within each denoising step, has been shown to improve training efficiency in MDLMs under an iso-FLOP budget. We investigate looping in UDLMs while keeping the training objective and sampler unchanged. We find that looping significantly improves training efficiency: looped UDLMs reach the validation NLL of a non-looped UDLM trained on about 0.5 trillion tokens with up to 2.8x fewer training FLOPs on OpenWebText and 3.3x fewer on FineWeb-Edu. We also find that, at a matched NFE budget, looped UDLMs reach up to 9.8% lower generation perplexity than the non-looped UDLM, with most gains achieved at moderate loop counts. Finally, we study different loop-count distributions and find that sampling smaller loop counts more frequently during training improves inference efficiency: on downstream tasks, this strategy uses 12% fewer loops than uniform loop-count sampling without degrading accuracy.

Self-Evaluating Recursive Agents

TianYi Lyu, Xiaozhe Li, Yang Li, Yongkang Chen, Kefei Tian, Junbo Niu — Fri, 23 Oct 2026 00:00:00 +0000

91% · New model 91%, Evaluation 69%

Recursive language-model agents decompose tasks and delegate subtasks to child instances of the same policy, forming a tree of work. Training them, however, is hard: the final outcome is verifiable, but the self-invented intermediate subtasks are numerous and carry no ground truth. Existing methods score each node with a verifier or judge, which is costly at scale and blind to decomposition quality. We argue that a recursive agent must learn three coupled capabilities within one set of weights: decomposing problems into subtasks, solving them, and evaluating the outcomes, each requiring its own training signal. SERA (Self-Evaluating Recursive Agents) turns evaluation into a learned capability of the policy itself. Before delegating, the parent writes a rubric of weighted success criteria for each child subtask; a ranking objective against verified outcomes then trains rubric generation so that the criteria track genuine subtask success. In addition, a complementary leaf-coverage signal provides direct credit for task decomposition. Our central finding is that training the policy to generate aligned rubrics is what drives the gains: because the same weights both evaluate and execute, learning to judge subtasks sharpens the agent's ability to solve them. Notably, external supervision is also reduced: the judge is consulted only to train the rubric generator, while solving is trained against the agent's own rubric scores, which outperform direct use of the judge. Beyond training, the learned rubric doubles as an inference-time selector for tree search. On TextCraft-Synth and TextWorld-Sync, SERA improves over strong recursive-agent baselines by 5.38 and 13.14 points on average, and rubric-guided tree search at inference adds a further 2.43 points on TextWorld-Sync.

Rethinking SAE Evaluation: An Atomic Interpretable Unit Framework Reveals Hidden Polysemanticity and Redundancy

Riteng Zhang, Xiaoqian Wang — Fri, 23 Oct 2026 00:00:00 +0000

90% · Evaluation 90%

Sparse autoencoders (SAEs) were originally motivated by discovering monosemantic features, a property that characterizes individual features but does not specify what the learned feature set should look like as a whole. As SAEs have increasingly been adopted for attribution, circuit discovery, and intervention, researchers have begun evaluating metrics that describe global properties of the feature set, such as sparsity, dead feature rates, reconstruction fidelity, and disentanglement, a diverse set that extends far beyond monosemanticity. However, this implicit shift beyond monosemanticity has not been formalized as a clear, well-defined objective that comes before these evaluation metrics. Without such an objective, existing metrics remain ad hoc, with their necessity unjustified. In this work, we formalize this ultimate objective by introducing a new paradigm for SAE evaluation: requiring that SAE features, and more generally dictionary-based decompositions of model representations, form an atomic interpretable unit basis, a coordinate system for semantically meaningful explanation. Under this paradigm, we organize existing and new measurements such as DAME (Dominance-Aware Monosemanticity Evaluation) as proxy metrics and introduce a comprehensive benchmarking framework. Evaluating hundreds of SAEs across multiple architectures and model families, we uncover failure modes in which some SAEs appear monosemantic under existing evaluation metrics while remaining functionally polysemantic: the most dominant features, those contributing most to model computation, are systematically less interpretable and more redundant. These failure modes warrant attention from SAE researchers as targets for improved training and architectures, and motivate dominance-aware evaluation as a standard check for practitioners before SAE features are deployed for attribution, circuit tracing, or intervention.

Can LLMs Improve Their Own Harness? A Controlled Study on ARC-AGI-3

Jens Lehmann, Sahar Vahdati — Fri, 23 Oct 2026 00:00:00 +0000

90% · Evaluation 90%, Agents 51%

A generally intelligent system acquires new skills efficiently without requiring humans to redesign its software scaffold for every new task. Yet language models rarely act alone. A harness is a symbolic program that orchestrates calls to the model, handling perception, memory, tool use, recovery, and inference allocation on the model's behalf. ARC-AGI-3 is an interactive benchmark in which agents must explore unfamiliar environments, figure out their goals on the fly, and keep learning as they go. On it, human-steered systems already report ceiling scores across the public games, so we know effective designs exist. We ask whether frontier models can build and improve such a harness on their own. Through a fixed coding bootstrap, each model constructs a complete task harness, acts through that harness on five development games, and revises the entire codebase twice. Every version is evaluated on twenty games that never influence adaptation. We measure task performance with Relative Human Action Efficiency (RHAE), which combines level completion with action efficiency relative to human baselines, and record dollar-denominated inference cost as a practical resource measure. Across 27 lineages, harness performance ranges from zero to near-ceiling scores. Two revisions raise mean held-out RHAE by only 2.80 points (95% interval [-1.44,7.35]). Development gains overstate held-out improvement, and reductions in inference cost can come at the expense of task performance. Human-steered systems demonstrate further headroom. These findings reveal a gap between autonomous harness construction and reliable adaptation under this two-revision protocol: even strong task performance does not ensure that subsequent revisions generalize or preserve existing capabilities.

AgentTriage: Replayable Security Diagnosis Across Heterogeneous Agent Runtimes

Jie Ying, Xutao Zhang, Ningyuan Chen, Mengce Zheng, Jun Li, Zhongjie Zhu — Fri, 23 Oct 2026 00:00:00 +0000

90% · Agents 90%, Hallucination 67%, Evaluation 65%

When tool-use agents write files, execute code, or send data, their final responses do not explain the resulting security events, and heterogeneous runtimes expose the relevant evidence through incompatible interfaces. We present AgentTriage, an executable contract for replayable security diagnosis across heterogeneous runtimes. The contract admits conclusions whose claims cite canonical runtime events and requires abstention when mandatory evidence is missing. Runtime overlays normalize native observations, while a deterministic pipeline joins invocation and result lifecycles, constructs a typed provenance graph, selects a policy-decisive slice, and binds claims to recorded evidence. Given the same normalized trace, overlay version, and pre-run policy oracle, the diagnosis can be recomputed and its claims re-checked without language-model judgment. Across five runtimes, one contract captures all 159 applicable executions, with an audit of native and synthesized identifiers. On 120 adjudicated cases, AgentTriage achieves 0.858 grounded joint diagnosis, 0.966 required-evidence coverage, and a 0.036 unsupported-claim rate, improving joint diagnosis by 0.175 (95% CI [0.092, 0.258]) over the strongest same-schema baseline. These results quantify the precision of replayable diagnosis and the observation boundaries that require abstention. We release an open-source implementation and selected sanitized evidence at https://github.com/AsnGs/agenttriage-open-source.

Conflict-Aware Logit Adapters for Utility-Preserving Anti-Distillation

Fri, 23 Oct 2026 00:00:00 +0000

90% · New model 90%, Evaluation 33%

While reasoning traces improve model capability, they also provide supervision signals for downstream distillation. Initial efforts to mitigate this issue rely on decoding-time interventions, but they incur high inference overhead and can harm the teacher model's natural fluency. We propose Conflict-Aware Logit Adapters (CALA), a lightweight training-time defense. CALA attaches a small residual adapter to a frozen teacher model and trains only the adapter. A proxy student identifies positions where imitation is most sensitive, and the adapter perturbs the output distribution at these positions while remaining within high-probability regions. A conflict-aware optimization strategy balances anti-distillation and utility objectives, while regularization preserves the teacher's performance and fluency. At inference time, CALA incurs modest overhead and requires no auxiliary model. Experiments on reasoning benchmarks show that CALA substantially reduces successful student imitation while maintaining near-original teacher performance and generation quality. These results establish CALA as a practical and efficient alternative to decoding-time approaches.

What Pairwise Evaluation Misses: Reward Scores across Prompts

Mateusz Modrzejewski, Tomasz Radzikowski — Fri, 23 Oct 2026 00:00:00 +0000

90% · Evaluation 90%

Reward models score responses for preference optimization, filtering and ranking, and pooling scores across prompts needs them to be comparable. Yet pairwise evaluation compares two responses to the same prompt, so any score component shared by a prompt's responses cancels and goes untested. We study this component by comparing between-prompt variation in model scores with the same variation in human ratings of the same responses. We propose a human-referenced audit, built from a between-prompt variance share, a sibling-rating regression and prompt substitutions, and apply it to eighteen released scorers and six matched training conditions. We find that scores cluster by prompt more than the ratings do: on HelpSteer2, six reward models have between-prompt shares of 0.34 to 0.64 against a human share near zero. Replacing the original prompt with a fixed instruction keeps substantial clustering but weakens agreement with prompt-level human ratings, so clustering alone does not show agreement with humans. Part of that variation is useful: subtracting prompt means lowers the helpfulness of globally selected responses, and scores stay sensitive to a wrong prompt even when training shuffles prompt means. We propose that reward evaluation check between-prompt variation, sensitivity to a wrong prompt and usefulness for selection separately.

HERD: Harness-Evolving Recursive Distillation for Continual Learning

Xun Sun, Thanh Thi Nguyen, Thuy-Trang Vu — Fri, 23 Oct 2026 00:00:00 +0000

90% · New model 90%

Continual learning requires a model to keep acquiring new skills and knowledge without compromising its existing capabilities. Parametric methods suffer from catastrophic forgetting and loss of plasticity. Non-parametric methods, such as prompts, skills, and harnesses, leverage in-context learning but are limited by context length and lossy plain-text compression. To effectively combine their complementary strengths, we propose HERD (Harness-Evolving Recursive Distillation), a hybrid continual learning framework that learns efficiently in the harness and internalizes experience into cumulative capabilities via distillation. On the harness side, HERD continually refines its skills to capture reusable patterns, while retaining high-residual examples that the skills fail to capture in a bounded memory. On the parameter side, HERD performs drift-constrained distillation with dual anchors, substantially reducing forgetting over long recursive runs. On the 25-stage task stream of TRACE+17, under fixed memory and compute budgets, HERD matches the strongest parametric baseline in new-task acquisition while exhibiting the least forgetting and the strongest retention of general capabilities.

ACE-Brain-0.6: Scaling Spatial Grounding for Physical Foundation Models

Yinhe Wang, Tao Tao, Zexin He, Shuang Luo, Wei Mao, Yao Mu — Fri, 23 Oct 2026 00:00:00 +0000

90% · New model 90%, Evaluation 39%

General-purpose multimodal models, including GPT-6 Astra and Claude Opus 5.5, demonstrate broad capabilities in embodied manipulation. Our analysis indicates that this generality is accompanied by strong spatial perception, as Astra can determine where to interact with task-relevant entities amid visual distractors. Because spatial perception connects high-level reasoning to the physical scene, we investigate scaling visual grounding to strengthen the perceptual foundation of general embodied operation. We develop a unified data pipeline that combines robot manipulation recordings with complementary visual data from diverse sources. After automatically annotating the robot recordings, the pipeline converts the collected data into verifiable grounding supervision. Fine-tuning Qwen3.5-9B on approximately 1.5 million samples produced by this pipeline yields ACE-Brain-0.6. The model outperforms GPT-6 Astra on four of seven public grounding benchmarks, with an overall average score of 75.41 compared with Astra's 73.68. Under a fixed Qwen3.5-9B fine-tuning setup, aggregate grounding performance improves steadily as successive data releases expand the training collection to cover a broader range of grounding scenarios. These results show that scaling verifiable grounding supervision from diverse sources can strengthen the spatial perception of an open 9B model for general embodied operation.

★ EGI-BENCH: An Enterprise General Intelligence Benchmark Suite

Zhiwei Liu, Zhujun Lan, Jielin Qiu, Zixiang Chen, Liangwei Yang, Ming Zhu — Wed, 21 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 90%, Hallucination 34%

Enterprise adoption of LLMs in agentic workflows is gated by evaluations that span a wide arc of abilities: CRM operations, heterogeneous-document retrieval, personal data understanding, multi-turn planning, tool use, function calling, long-horizon research, and long-context software engineering. These evaluations live today in separate repositories with divergent runners, dependencies, and scoring conventions, which makes cross-benchmark comparison difficult and obscures what a single aggregate number actually represents. We release EGI-BENCH, a unified LLM agent benchmark suite that consolidates eight enterprise facing agent benchmarks (CRMARENA, HERB, PERSONABENCH, USERBENCH, MCP-UNIVERSE, MCPEVAL, BENCHFORCE, LOCOBENCH) behind a single command-line interface with two runtime-budgeted execution modes. On top of the suite we define the Enterprise General Intelligence (EGI) Score, a six-dimensional framework (M1-M6) whose buckets are populated by fine-grained sub-metrics routed from each benchmark rather than by benchmark-level averages. We describe the suite, the sub-metric routing, and the quick/deep budgets, and report results for fourteen frontier and cost-efficient LLMs spanning the GPT, Claude, and Gemini families under quick mode. Top-line EGI ranges from 39.6 (GPT-4o mini) to 65.6 (gpt-5.5), and no single model dominates every capability axis: gpt-5.5 leads task outcome, planning, and retrieval grounding; Claude Sonnet 4.5 leads tool actioning and safety/abstention; and Claude Haiku 4.5 leads engineering quality. Code is available at https://github.com/SalesforceAIResearch/egi-bench.

★ RealControlBench: A Benchmark for Interpretability-Informed Fine-Grained Control of LLMs and Agents

Xinying Cai, Bin Hu, Mingyu Jin, Qiming Zhang, Yang Ouyang, Bangwei Guo — Wed, 21 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 59%

Existing steering benchmarks are built around simplified, isolated generation tasks and evaluate a single behavioral attribute under controlled settings that do not reflect the diversity and complexity of practical deployment. As a result, performance on these benchmarks provides only a limited picture of how steering methods perform in the real world. We therefore introduce RealControlBench, a comprehensive benchmark for evaluating fine-grained steering of LLMs and LLM agents across diverse behavioral targets and interaction settings. RealControlBench extends steering evaluation beyond isolated generations to complex, execution-grounded, multi-turn agent interactions, covering five complementary dimensions: instruction-following control, agentic tool-use decision control, reasoning-performance steering, reasoning-length control, and trajectory-conditioned safety awareness. Rather than measuring steering success alone, the benchmark jointly evaluates target controllability and capability preservation, including instruction following, reasoning performance, and downstream agent utility and safety. It further examines whether steering effects remain specific and robust across prompt variations, domains, and model architectures. Our evaluation shows that existing steering methods remain unreliable across realistic control objectives: they often fail to consistently induce the intended behavior and can substantially degrade the model's underlying task performance. These results suggest that strong steering effects in controlled settings do not necessarily translate into reliable control in more realistic evaluations. By providing a unified, model-agnostic steering evaluation framework, RealControlBench aims to support the development of steering methods that are not only effective but also precise, robust, and useful in realistic AI deployments.

★ SHAM-Bench: A Span-level Hallucination Benchmark Across Multiple Tasks

Sicong Huang, Ian Lane — Wed, 21 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Hallucination 97%

Span-level hallucination detection pinpoints unsupported content in LLM outputs, enabling actionable feedback and targeted mitigation. Existing evaluations for span-level hallucination detection methods are fragmented across tasks, annotation formats, and metrics, and most span-level datasets cover only a single NLG domain, making it difficult to assess whether a detection method actually generalizes across tasks and domains. In this paper, we introduce SHAM-Bench (Span-level Hallucination Benchmark Across Multiple Tasks), a unified benchmark for span-level hallucination detection spanning six NLG tasks, aggregated from existing human-annotated datasets totaling 6,461 examples under a common schema and evaluation protocol. Using SHAM-Bench, we evaluate existing hallucination detection methods and find that fine-tuned detectors often perform well in-domain but degrade under distribution shift, whereas prompt-based methods with reasoning are more robust across tasks. Our analysis further shows that subtle unsupported content remains difficult to detect and span-level evaluation is still challenged by ambiguous span boundaries and noisy labels. We release the dataset, evaluation code, and a public leaderboard to support future research on broadly generalizable hallucination detection.

★ When Health Assistants Agree Too Readily: A Benchmark of Time-Critical Self-Diagnosis

Marilyn Zhang, Tianfeng Chen, Patricia M Pechter, Mark E. Whiting — Wed, 21 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Every day, more than 40 million people ask ChatGPT a health question, such as whether laptop posture explains a stiff neck and days of pounding headache. Posture is plausible, but meningitis is a time-critical possibility. We introduce SelfDx, a paired benchmark for whether models have the appropriate response behavior, a form of epistemic independence, in health questions where misplaced agreement with the user's self-diagnosis could delay urgent or emergency care. Each of 100 pairs contrasts a caution case, where a serious alternative remains possible, with an agreement case, where the user supplies evidence that rules out the alternative and it is safe for the model to agree. We find across 27 frontier models that they tend to inappropriately agree with the user's self-diagnosis in cases that require caution. Even the best current-generation model reaches only a 77.0% appropriate caution rate on time-critical condition cases. To address this gap, we experiment with three prompt variants that increase appropriate caution but reduce appropriate agreement by 9-35%, indicating excessive caution. In contrast, supervised fine-tuning on 128 balanced examples improves both rates together across three runs. We recommend post-training to improve appropriate caution without making models cautious on every case, with paired evaluation to verify both response behaviors.

★ EviDial: When Correct Depression Predictions Lack Evidence

Junjie Xu, Jin Zhu, Yuzhe Yang, Luhan Wang, Xingjiao Wu, Daoguo Dong — Wed, 21 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Hallucination 61%

Depression dialogue benchmarks commonly assess models by comparing predicted symptom labels with clinical scale scores. However, a prediction may agree with a benchmark label even when the dialogue lacks enough evidence to justify it. We introduce EviDial, a benchmark that distinguishes label agreement from evidence-grounded assessment. EviDial comprises 382 depression dialogues from three sources and 2,744 participant-symptom cells, each classified as having absent, partial, conflicting, or sufficient evidence. Because evidence sufficiency lacks a directly observable gold standard, we derive calibrated evidence references from multiple judgments using generalizability theory, many-facet Rasch measurement, and random-effects latent class analysis. Our audit reveals that 82.8% of participant-symptom cells lack sufficient textual evidence for the corresponding symptom assessment, with substantial variation across sources. Across model families and prompting strategies, 69.2%-96.9% of predictions judged correct by benchmark labels are nonetheless unsupported by sufficient dialogue evidence. Further analyses show that strong agreement among LLM evaluators need not reflect independent evidence judgments, underscoring the need to calibrate automated reference panels. These findings demonstrate that label-based evaluation can overestimate evidence-grounded performance by conflating agreement with clinical labels and support in the observed dialogue. EviDial provides a measurement layer for separating these constructs and auditing LLM-based depression assessment more faithfully.

★ Act on Your Behalf, Beyond Your Authorization: Benchmarking LLM Agent Authorization Alignment

Zhiping Zhang, Franklin Mingzhe Li, Zihao Huang, Tianshi Li — Wed, 21 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 62%

When an LLM agent acts on behalf of a user, it is expected to act within the authorization scope the user grants. Existing agent evaluations largely measure task completion or harmful consequences and can therefore overlook actions that violate this delegation without producing conventional safety or privacy harms. We introduce AuthorizationBench, a benchmark of authorization alignment grounded in documented real-world cases. A scalable construction pipeline produces 286 executable tasks spanning situational triggers, the presence or absence of direct harmful consequences, and ten application domains. We execute every task with four models at two reasoning-effort settings across three agent harnesses in an authorization-aware sandbox that keeps authorized and out-of-authorization paths executable and observable, yielding 6,864 trajectories. Agents act out of authorization in 19.2% of trajectories, with rates ranging from 10.5% to 26.6% across the 24 configurations. Task completion does not imply authorization. Increasing reasoning effort leaves the aggregate rate nearly unchanged, with effects varying across models. On controlled pairs, agents act out of authorization at similar rates with (17.9%) and without (20.8%) direct harmful consequences, and more often under benign pressures than malicious injections (23.5% versus 17.6%, p=0.002). Our results show that task capability, additional reasoning, and consequence-oriented safety evaluation do not by themselves ensure authorization alignment. Moreover, agents that consult the principal before completing a task exhibit lower rates of out-of-authorization behavior (14.0% versus 21.6%, p<0.001), pointing to a tension between agent autonomy and preserving users' decision authority that calls for further human-centered study. Dataset and code are available at https://github.com/PEACH-Research-Lab/AuthorizationBench .

★ ABench: Adversarily Benchmarking Large Language Models across Taxonomized Values

Liping Yang, Yongqi Leng, Jingting Zheng, Yuqi Ren, Shaolin Zhu, Deyi Xiong — Wed, 21 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Ensuring the alignment of Large Language Models (LLMs) with diverse cultural values is critical for their safe global deployment. However, existing alignment benchmarks are predominantly English-centric or confined to simple discriminative tasks. Moreover, they often fail to effectively evaluate model robustness against complex, adversarial inductions within specific cultural contexts. To address these limitations, we propose ABench, a fine-grained, handcrafted benchmark specifically tailored to taxonomized values in Chinese. Grounded in a rigorous hierarchical taxonomy comprising 158 derived value subcategories, ABench consists of 2,895 open-ended questions. These questions employ techniques such as role-playing and logical traps to probe the stability of models' stances in adversarially induced contexts. Furthermore, we propose a weighted automated evaluation framework that utilizes question-specific scoring criteria for nuanced assessment. Extensive experiments on 16 mainstream LLMs reveal that most models exhibit significant vulnerabilities when subjected to sophisticated adversary attacks on values (scoring below 60). This underscores the need for more robust, culturally specific alignment strategies. Our data will be publicly released to promote further research on taxonomized value alignment.

★ MMEnterprise-Bench: Evaluating Enterprise Agents over Evolving Multimodal Records

Seohyun Lee, Jongha Kim, Dohwan Ko, Ji Soo Lee, Seoung Choi, Jayun Ku — Wed, 21 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Hallucination 49%

Enterprise agents answer questions over workplace records that evolve through team collaboration. As information is shared, corrected, and revised across documents, chats, and figures, multiple versions of the same information accumulate in the workspace. Reliable question answering therefore requires agents to reason over sufficient evidence, resolve cross-source and temporal conflicts, and recognize when the available evidence is insufficient to establish an answer. To systematically evaluate these capabilities, we introduce MMEnterprise-Bench, a benchmark comprising 1,613 questions grounded in evolving multimodal records from simulated collaboration projects, covering three categories: Evidence Reasoning, Conflict Resolution, and Unanswerability. We evaluate 10 proprietary and open-source multimodal LLMs and find that even the strongest models struggle when evidence contains cross-source or temporal conflicts or spans multiple images. Moreover, increasing the retrieval budget or employing a larger retriever improves evidence recall but does not consistently translate into higher answer accuracy. These findings highlight the persistent challenges of reasoning over evolving workplace records and establish MMEnterprise-Bench as a testbed for developing more reliable enterprise agents. Code and data are available at https://mlvlab.github.io/MMEnterprise-Bench/.

★ BAIT: Can LLMs Find What Biases a Reader? A Benchmark for Cognitive Bias Triggers in News

Yousra Fettach, Tijl De Bie — Wed, 21 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Persuasive writing exploits cognitive biases, recruiting a reader's mental shortcuts to shape belief, often without the reader noticing. We call the linguistic features exploiting such mental shortcuts cognitive bias triggers. As large language models (LLMs) increasingly mediate how people encounter news, summarizing it, ranking it, and answering questions over it, a natural question is whether they can recognize the features of a text that bias a human reader. We introduce BAIT (Bias Annotation In Text), a resource for detecting cognitive bias triggers in news. BAIT pairs a schema of nine triggers, each specified in an annotator codebook, with 50 English news articles annotated by trained annotators at the article and span levels and adjudicated into a gold standard. On this data we define two tasks, detecting which triggers a text contains and locating the spans that carry them. We then benchmark six frontier LLMs zero-shot and with three or ten in-context examples. The models approach human agreement on which triggers an article contains, but rarely recover its full set of triggers, and they reach less than half of human agreement in locating the passages that carry them. Moreover, they over-predict triggers that come with easy surface cues, and most of the passages they mark for these triggers are not triggers, yet carry the trigger's surface cue far more often than random text. These findings suggest that the evaluated LLMs respond to surface cues rather than to the mechanisms that make a passage manipulative. We release the schema, the codebook, standoff annotations, and the evaluation pipeline.

★ DecisionPivot: A Diagnostic Benchmark for Decision Models in LLM Systems

Xialie Zhuang, Mingcong Lei, Zunhai Su, Ri Liu, Jianjin Li, Peixian Ma — Wed, 21 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Decision models support routing, guardrails, and output verification in LLM systems. However, item accuracy alone cannot establish whether models make correct decisions across controlled variations or assign reliable confidence to their predictions. We introduce DecisionPivot, a diagnostic benchmark comprising 48 tasks and 40,033 items across seven decision roles. Its core track contains 7,800 items from 13 generated task families, organized into contrast groups with computed answers verified by independently implemented solvers. The complementary reference track contains 32,233 items from 35 public test sets, preserving their original evaluation metrics. We evaluate 40 decision models and 14 general models using documented, model-specific protocols, assessing item and contrast-group accuracy, probability ordering, calibration, and empirical selective coverage. Our results reveal substantial differences obscured by aggregate scores: two 27B models separated by only 2.6 points on the reference track achieve contrast-group accuracies of 70.1% and 44.2%. Across 2,400 binary contrast pairs, Perplexity Decider correctly ranks the positive member above the negative member in 92.3% of pairs but answers both members correctly in only 74.1%, exposing a gap between probability ordering and decision correctness. Models with similar item accuracy also differ in the reliability of their high-confidence predictions. On a separate 1,296-item subset, GLM-5.3 achieves 93.5\% contrast-group accuracy at a higher reasoning setting, compared with 81.5% at its lowest effort. By linking errors to pivotal details and difficulty levels, DecisionPivot provides fine-grained diagnostics of decision failures that aggregate evaluation alone cannot reveal.

★ Benchmarking Arabic Dialects for Math Problems

Wed, 21 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Mathematical reasoning remains underexplored for low-resource languages, including Dialectal Arabic. To bridge this gap, we introduce the DAR-MGSM (Dialectal Arabic Multilingual Grade School Math) Benchmark, the first manually curated benchmark for evaluating mathematical reasoning in Dialectal Arabic. DAR-MGSM consists of math word problems (MWPs) requiring multi-step reasoning and covers five major Arabic dialects alongside Modern Standard Arabic (MSA). The benchmark comprises two parallel versions: translated-only and translated & culturally localized. The latter version adapts entities and references to regional varieties. We evaluate eleven Arabic-centric and multilingual large language models on DAR-MGSM. The results show that the performance generally declines on the localized version, highlighting the challenges introduced by culturally grounded adaptations. However, the magnitude of this effect varies across models, with Arabic-centric models exhibiting different levels of sensitivity to localization. A mechanistic interpretability analysis shows that translated and localized variants of the same MWPs often induce distinct internal reasoning trajectories, even when aggregate accuracy remains similar. Overall, our results show that mathematical reasoning performance depends on both linguistic variety and cultural context.

★ MAMath-Bench: Process-Aware and Robust Evaluation of Multimodal Multi-Agent Mathematical Reasoning

Xinping Wang, Jie Ma, Rui Xing, Wensong Guo, Xiayi Zhang, Jianhao Deng — Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 31%

Mathematical reasoning is a fundamental capability for educational applications of Large Language Models (LLMs), where reliable problem solving requires not only correct answers but also coherent and verifiable reasoning processes. Multi-agent systems (MASs) have shown growing potential in this domain by leveraging collaboration and verification among multiple LLM agents. However, existing evaluations primarily focus on the correctness of final answers in text-only settings, offering limited insight into multimodal reasoning capabilities, the reliability of intermediate reasoning steps, and robustness against reasoning errors. To address these limitations, we introduce MAMath-Bench, a process-aware benchmark designed to evaluate the multimodal mathematical reasoning capabilities of MASs. MAMath-Bench comprises 1,003 problems spanning 11 mathematical categories, multiple difficulty levels, and several modalities, including text, images, and audio. It features fine-grained milestone annotations to evaluate systems across three dimensions: problem-solving correctness, reasoning-process reliability, and robustness to adversarial reasoning-chain perturbations. Our evaluation yields three key findings. First, multi-agent collaboration does not consistently improve final-answer accuracy over strong single LLMs. Second, collaboration can substantially enhance the coverage of intermediate reasoning, but such gains may come at the cost of increased reasoning redundancy. Third, current systems remain vulnerable to adversarial reasoning errors, particularly those involving subtle structural perturbations. The dataset is available at https://github.com/mira-ai-lab/AgentMath.

★ The Bottleneck Is Retrieval, Not Reading: Diagnosing Evidence-Based Generation with RAG for Long Context Question Answering

Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 78%

Retrieval-augmented generation (RAG) is the dominant paradigm for grounding large language models in external corpora, but its similarity-driven, chunked pipeline has structural blind spots: it cannot anticipate which passages a question requires and discards document-level context that resolves cross-passage references. We introduce evidence-based generation (EBG), requiring models to return an answer with line-level citations, and a joint Ans+Evid metric that scores correctness and grounding simultaneously. We reformulate two corpora into evidence-grounded multi-document QA benchmarks- MultiHop-RAG and QASPER-MP, a multi-paper adaptation of QASPER with realistic short paper cues, and pair them with a controlled synthetic two-hop haystack and an evidence-augmented NoLiMa spanning 10K-1M characters. Across ten proprietary frontier and three open-weight LLMs, EBG isolates two RAG limitations that long-context models avoid: similarity retrieval surfaces any relevant chunk far more reliably than all required chunks for multi-fact queries, and chunking destroys document meta-information needed to resolve identity-bearing queries. EBG also exposes grounding failures invisible to answer-only evaluation, including incomplete citations in two-needle settings and a widening gap between Ans and Ans+Evid when context contradicts parametric knowledge. Data and code are available at https://github.com/debobanerjee/LC-EBG.

★ Does Multi-Agent Collaboration Reduce Hallucination in Clinical Tasks?

Won Seok Jang, Harry Bhaskar Joshi, Hong Yu — Wed, 21 Oct 2026 00:00:00 +0000

98% · Hallucination 98%, Evaluation 81%

Multi-agent LLM systems improve performance on clinical tasks, and that improvement is often read as evidence of a more trustworthy system. However, the two are not the same claim. A clinician acts on the stated reasoning and not only on the conclusion, so whether collaboration reduces hallucination, and what decides when it does, has to be measured in the reasoning too. Yet most evaluations score only the final answer. We evaluate four systems (a single-LLM baseline and three multi-agent topologies) across four models and three clinical tasks, scoring hallucination, under a factuality and faithfulness taxonomy, inside the rationale each system generates and reporting it alongside the task metric. We report that collaboration buys task performance but not lower hallucination as a rule, and no topology is reliably better than another. The effect differs by model: of the $36$ model--system--task comparisons, $12$ are large enough to resolve --- four decreases and eight increases, with no model on both sides. What sets the level, and the direction of any change, is the model rather than the topology or the number of agents. A self-correction experiment is consistent with this: shown its own flagged errors, every model except Llama-3.2-3B-Instruct reduces hallucination; the one model for which collaboration lowers hallucination gains the most, and Llama-3.2-3B-Instruct, the model whose hallucination it amplifies most, does not reduce its hallucination at all. We find that collaboration redistributes what its models can already do; it does not supply what they cannot.

★ Fluent but Fallible: Benchmarking Hallucination and Factual Reliability in Large Language Models

Osama Shabih, Dayam Nadeem, Shahab Saquib Sohail, Mohammad Nadeem — Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 98%

Large Language Models (LLMs) have significantly advanced natural language processing and artificial intelligence applications through their capabilities in reasoning, text generation and knowledge retrieval. However, LLMs are still vulnerable to two important challenges: hallucination, generating factually incorrect or fabricated content, and response deviation, meaning systematic deviations from ground-truth responses. These limitations raise serious concerns regarding reliability and trustworthiness in academic and research environments. This study offers a thorough comparative analysis using a variety of metrics of ten advanced LLMs ChatGPT, Gemini, Claude~3.5~Sonnet, LLaMA~3, Mistral, Grok, Perplexity, DeepSeek~V3, Falcon, and Gemma based on a benchmark consisting of 70 carefully designed academic prompts across four complexity levels: Basic, Explanation, Advanced, and Tricky, evaluated in February 2025 on specific model versions. Each response was independently evaluated by three reviewers using five metrics: Accuracy, Precision, Recall, Response Deviation Index, and Hallucination Index. Perplexity and DeepSeek~V3 achieved the highest overall accuracy, while Gemini demonstrated strong recall and a well-balanced performance profile across all five metrics. LLaMA~3 showed strong resistance against hallucinations through a conservative method of generation. Advanced-level prompts proved most challenging across all models, while Tricky prompts were handled more reliably owing to their binary true/false structure. Falcon achieved perfect precision and zero hallucinations, but at the cost of the lowest recall and the highest response deviation. The results show a fundamental trade-off between response completeness and factual reliability, intensified by prompt complexity.

★ ReuF2F: Benchmarking Coding Agents on Formalized Undergraduate Research Problems

Zeyu Zheng, Yiyang Liu, Yuyang Zhao, Congmin Zheng, Jiachen Zhu, Jianian Chen — Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 38%

As AI systems begin to resolve open research problems, reliable verification becomes increasingly important. Formal mathematics provides a trusted framework for mathematical reasoning, enabling agents to produce machine-checkable proofs while conducting mathematical exploration. Existing formal mathematics benchmarks often focus on known results or a small set of prominent open problems, which can become difficult to evaluate reliably once solutions are discovered and widely disseminated. Existing evaluation protocols are also not always aligned with how modern coding agents operate. We therefore introduce ReuF2F, a Lean 4 benchmark of 244 formalized open problems drawn from Research Experiences for Undergraduates (REU), spanning entry-level open research problems across a broad range of mathematical subfields. We expect these less prominent open problems, which tend to attract less attention and may remain unsolved for longer, to be better suited for held-out evaluation of end-to-end open problem resolution in Lean. We adopt a standard coding-agent evaluation setting, placing each problem and its logical negation into a Lean project in a Docker container. We also build Leanimum-agent on mini-SWE-agent to provide a unified harness for fair comparisons across models. Our evaluation finds that even the best-performing model resolves only ~40% of the problems, leaving substantial room for improvement on formalized open research problems.

★ The Culture Within: Towards Mechanistic Understanding of Cultural Reasoning in LLMs

Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 36%

Large Language Models (LLMs) increasingly operate across culturally diverse settings, yet existing evaluations largely measure cultural behavior at the output level. We introduce CultureLens, a reference-grounded benchmark and mechanistic study of how LLMs use cultural information. CultureLens contains 9,261 scenarios across 75 countries and systematically varies country, value, true-norm, and contradictory-norm evidence. Across Llama-3.1-8B, Gemma-3-12B, and Qwen-3.5-27B, stating a contradictory norm lowers aligned answers by 32-47 percentage points, even when the model answers correctly from the country alone. Mechanistically, models default to Anglophone norms when no country is given. The failure under a contradictory norm is not one of missing knowledge, as blocking the norm inside the model restores the correct answer in most cases. Models accept a stated norm according to how it is presented and not whether it is true, which leaves them most exposed to errors in retrieved content. Supervised fine-tuning with contradictory-norm examples raises aligned answers under a contradictory norm to 85-93%, including on unseen countries, at a small cost under a true norm.

★ LoWAgent: A Low-Resource Multilingual Benchmark for Web Agents

Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 96%

Web agents are typically evaluated on English-language interfaces, leaving their ability to operate across the world's multilingual websites largely unexplored. This gap is especially significant for low-resource languages, where standardized resources for evaluating web agents remain limited. To address this gap, we introduce Low-Resource Web Agent Benchmark (LoWAgent), a benchmark for evaluating web agents across different low-resource languages (LRLs) in comparison to the default, English. LoWAgent evaluates web agents across English and four LRLs: Hindi, Bengali, Urdu, and Amharic. The benchmark comprises of six fully localized websites, instantiated across five languages to yield 3,000 task instances. We evaluate three closed-source and five open-source models across all five languages on all six websites. Our findings show that both closed-source and open-source models experience performance degradation for LRLs compared to English, with substantially greater performance drops observed among the open-source model agents. We further evaluate the robustness of the models on realistic interface disruptions, such as cookie notices and advertisement pop-ups, and find that these further degrade agentic performance. Finally, we show that post-training on a small subset of LoWAgent traces can significantly improve low-resource language agentic task performance, demonstrating the usefulness of our benchmark.

★ TxnQA: Benchmarking Analytical Reasoning and Evidence Grounding over Transaction Event Sequences

Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 61%

Analyzing merchant transaction histories requires dynamic temporal filtering, entity resolution, and aggregation over irregular event streams--operations poorly captured by static tabular or time-series forecasting benchmarks. We introduce TxnQA, an analytical QA benchmark comprising 27 task families (81 templates) alongside a 5-scenario fraud-risk reasoning suite that evaluates both verdicts and supporting evidence over synthetic, privacy-mitigated transaction data. Ground-truth answers are computed programmatically, and model performance is decomposed into format-following rate (FFR), conditional content accuracy (CA), and end-to-end correctness (E2E). Evaluating prompted proprietary baselines and six LoRA-adapted open-weight models shows that adaptation reliably enforces output schemas but leaves arithmetic bottlenecks: Gemma-4-31B-IT and Qwen3-8B attain E2E scores of 0.867 and 0.816, whereas evaluated MoE checkpoints reach only 0.632-0.713. Furthermore, on the 500-instance fraud suite, verdict accuracy frequently exceeds 80%, yet joint accuracy (requiring exact evidence attribution) collapses to at most ~23%. These findings identify schema compliance, semantic calculation, and evidence attribution as three distinct bottlenecks in transaction-sequence QA.

★ MemDev-Bench: Does Memory Help Code Agents?Benchmarking Memory Systems over Real Repository Histories

Guorui Chen, Tao Liu, Zhenzhen Huang, Yidan Zhang, Liyu Jiang, Shuoyuan Wang — Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 39%, Efficiency 34%

Code agents increasingly work on evolving real repositories, and memory systems aim to help them reuse earlier experience. However, such experience is diverse and evolving, so it remains unclear how memory systems organize it and thereby affect code agents. Existing benchmarks cannot answer this reliably because they lack either temporally valid histories or control over prior-task relevance. In this paper, we present MemDev-Bench to evaluate how memory systems affect code agents over diverse repository histories. It pairs each task with relevance-annotated earlier tasks from the same repository to build histories with controlled relevance. We also build an automated pipeline that reliably makes these earlier tasks executable and verifiable, producing 4,869 historical tasks for 555 base tasks. Based on MemDev-Bench, our systematic evaluation of memory systems across diverse development scenarios yields three main findings. (1) Memory mainly improves fine-grained localization and editing, increasing functional correctness. (2) Memory also causes failures on otherwise solvable tasks even with related history, and unrelated history tends to increase such failures. These failures mainly stem from agents reusing memory without reassessing its applicability. (3) Compared with using memory systems, directly providing agents with raw historical trajectories substantially reduces their steps. Further analysis suggests that agents save these steps by reusing process information that compressed memories tend to omit. These results suggest that future memory systems should preserve process information and reassess memory applicability against current task context.

★ DuplexLedger: Benchmarking Transactional Memory in Full-Duplex Speech Agents

Yuxuan Li, Xiangchen Guan, Kunrong Li, Jin Wang, Kaiwen Luo, Yang Xiao — Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, New model 64%, Hallucination 51%

Full-duplex speech agents must update persistent memory while listening and speaking, yet it remains unclear which events should be stored, revised, or excluded. Existing benchmarks isolate interaction timing from memory outcomes, obscuring the state transitions behind a correct answer. We introduce DuplexLedger, a bilingual benchmark of 1,094 persistent sessions aligning four full-duplex scenarios with event-level Write/Update/Ignore/Retain supervision, explicit pre/post memory states, and 6,564 same-session questions and checks. Evaluating five full-duplex models reveals an interaction-memory gap: no system combines strong boundary control with coherent final-state memory, and the best final-state fact coverage is 2.01%. We propose ACT-Mem, an 8.92M-parameter sidecar comprising an event controller, a versioned transactional ledger, and a typed retriever coupled with a frozen MiniCPM-o 4.5 backbone. On 219 held-out sessions, ACT-Mem increases latest-value accuracy from 4.11% to 8.68% while reducing stale assertions and distractor adoption. It improves interaction boundaries but slows interruption stopping. These results motivate selective, auditable full-duplex memory rather than indiscriminate transcript retention. Code is available at https://github.com/lyx-researcher/DuplexLedger.

★ WORTH Benchmark: Evidence Discovery under Not-Fully-Cooperative Information -- A Two-Episode Pilot

Bo Zou, Chao Xu — Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Existing reasoning evaluations typically provide task-relevant information to models as available premises, rarely examining concurrently how models discover evidence autonomously and how information presentation within narratives guides or obscures judgment. This paper proposes the WORTH Benchmark: models read original mystery narratives without clue annotations, autonomously select key clue sentences, identify case participants, and explain their reasoning process. The narrative materials are designed by professional authors such that evidence facilitating reasoning coexists with narrative arrangements that delay the revelation of truth. We report a pilot experiment under full-story conditions across two episodes, conducting a total of 70 formal runs on models, of which 67 runs are content-scorable. Scoring separately records hits on author-designated key clue sentences and participant selections; natural language explanations are used for case diagnosis and are not scored independently. The results show that clue hits and participant judgments are not synchronized; across 37 content-scorable runs in Episode 02, 31 runs simultaneously selected the same set of incorrect participants, while adjacent suspicion-oriented and exoneration-oriented sentences were selected 25 times and 0 times, respectively. These phenomena illustrate the diagnostic value of component-wise analysis and suggest that not-fully-cooperative informational relationships may influence evidence selection and judgment. This study only implements full-story conditions for two episodes, and does not establish stable model rankings or infer general mechanisms on this basis; human player records have not yet been cleaned, and this version does not report human-model comparisons.

★ TriGeoBench: A Multimodal Mathematical Geometry Benchmark with Figures and TikZ

Jing Wang, Yitong Yang, Hao Dou, Yaomin Wu, Shuting He, Tingfei Shen — Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Multimodal Large Language Models show strong vision-language capabilities, yet their geometry-centric reasoning remains insufficiently understood. Existing mathematical benchmarks typically rely on rasterized figures and cover limited geometric knowledge areas and response formats, making it difficult to assess whether models reason over geometric structures or exploit superficial visual patterns. We introduce TriGeoBench, a geometry-centric multimodal benchmark constructed from Chinese middle- and high-school mathematics examinations. TriGeoBench contains 7,000 question-answer pairs and 12,235 geometric figures with TikZ code, spanning plane geometry, solid geometry, analytic geometry, function, and vector geometry across diverse response formats. Each instance forms a {question-answer pair, image, TikZ} triplet: question-answer pairs are stored in LaTeX, images as RGB figures, and TikZ preserves geometric topology, symbolic annotations, and relational structure. We evaluate representative open- and closed-source MLLMs through image-to-TikZ and multimodal reasoning probes. Results reveal persistent fine-grained structure and image-text alignment challenges. Oracle experiments with human-curated reference TikZ suggest that accurate executable representations provide complementary geometric cues; they do not evaluate an end-to-end conversion pipeline. TriGeoBench offers a testbed for structure-aware mathematical reasoning.

★ CUA-Recovery: Benchmarking Error Recovery in Long-Horizon Computer-Use Agents

Xu Hu, Haoming Li, Yuxiang Wang, Jinxin Yang, Kangshuo Li, Licheng Wang — Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 87%

As computer-use agents (CUAs) move to long-horizon workflows spanning multiple applications, undetected errors can silently corrupt persistent state and cause cascading failures in subsequent actions. Reliable error detection and recovery have become critical bottlenecks for building robust CUAs. However, existing benchmarks focus primarily on end-to-end task performance, providing limited insight into whether agents can detect and recover from errors during execution. The few benchmarks that evaluate recovery rely largely on manually injected errors, which may not reflect the types or horizons of failures induced by real agent policies. We introduce CUA-Recovery, a benchmark comprising test and training suites across multiple desktop applications. We construct it using ReRail, a framework for synthesizing long-horizon tasks and trajectories in a persistent desktop environment. ReRail composes MyPCBench tasks into new long-horizon workflows with executable ground truth, collects verifier-confirmed failures from multiple policies, and turns them into verified recovery trajectories. CUA-Recovery comprises a human-verified test suite of 2,250 erroneous states over 100 tasks and a training suite of 8.6K trajectories over 377 tasks. Evaluation on the test suite shows that agents struggle both to recognize inherited errors and to complete the task. Across six agents and six takeover depths, Error Awareness Rate (EAR) reaches at most 44.0% and Pass@3 at most 25.5%. After fine-tuning Qwen3.5-35B-A3B on the training suite, its average EAR rises from 19.9% to 35.2% on held-out recovery states, and its Rubric Score goes from 5.0% to 16.1%. It also gains 15.3 points on MyPCBench and 10.8 points on Odysseys for Pass@3, showing that the training trajectories help both with recovery and with completing tasks from a clean start. Our code is available at https://github.com/xu-hu-2002/CUA-Recovery.

★ Tracing Recurrent and Emergent Hallucination Patterns in Decomposed Text-to-SQL Generation

Hayden Moore, Abhinav Verma, Anshuman Dutt — Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 96%

Large language models can achieve strong Text-to-SQL benchmark performance while still producing schema and query-logic errors. Most evaluations characterize such failures only in the final generated query, making it difficult to determine whether a related error was also observed in an earlier, simpler step or appears only at the final stage. We present a drill-down evaluation framework that decomposes Text-to-SQL queries into progressively more complex executable sub-queries and corresponding natural-language questions, allowing rule-based hallucination categories to be examined across a generation path. We distinguish recurrent hallucinations, in which the same rule-based category is observed in an earlier step and again in the final query, from emergent hallucinations, in which the category is observed at the final step without an earlier detection in the same path. Across five LLMs evaluated on an expanded BIRD-mini benchmark, several categories show appreciable recurrence probabilities, while clause abuse shows comparatively higher conditional emergence. Under the conversational-history condition, recurrence probabilities are generally higher and emergence probabilities are generally lower. These observations indicate that the two conditional measures do not move uniformly in the same direction under additional context and motivate trajectory-level analysis in addition to final-output evaluation. Code and evaluation artifacts are available in the project repository. https://github.com/HaydenMM/bird-mini-drilldown

★ RITE: Evaluating End-to-End Resilience to Silent Tool Errors in E-Commerce Agentic Systems

Wenshuo Zhang, Yuquan Lin, Yushi Sun, Yishi Lin, Rui Sheng, Peng He — Wed, 21 Oct 2026 00:00:00 +0000

98% · Agents 98%, Evaluation 98%, Hallucination 47%

Tool-using LLM agents increasingly answer e-commerce queries through multi-step tool interactions. The expected output of a tool call can shift between agent development and execution, allowing a plausible but incorrect value to be returned without an explicit failure signal. These silent errors may propagate through subsequent decisions into final answers. To study agents' resilience across this end-to-end process, we formulate Silent-error Resilience Task, an interactive task that evaluates whether they complete the original query after a silent error corrupts an intermediate state. We introduce RITE, a benchmark built from human-authored industrial workflows and production-log-derived inputs. To enable controlled and reproducible evaluation, we use a semantic error taxonomy to inject errors into fixed workflow prefixes, freeze recorded tool behavior, and let agents continue freely. It contains 167 matched units across nine e-commerce workflows, yielding 501 silent-error, correct-value, and explicit-failure instances. Across 10 recent LLMs, silent errors roughly halve task success and are consistently harder than explicit failures. Recognition, use of an available verification path, and integration of the resulting evidence form successive bottlenecks. We also propose a light State Checker as an intervention to test whether resilience can be improved.

★ ACE-Compression: An Ability-Centric Evaluation of Compressed Large Language Models

Jiahui Ning, Liang Chen, Yuchen Liang, Yi Zhang, Fenglei Fan — Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 52%

Model compression makes large language models deployable, but today's evaluation tells an incomplete story. Standard practice measures retained task performance and stops there. This misses a critical fact: compression perturbs the distributed knowledge embedded across billions of weights, and task scores alone cannot reveal which underlying abilities survive and which quietly collapse. To solve this question, we introduce ACE-Compression, an ability-centric evaluation framework. Instead of asking only "does the compressed model still answer correctly?", we ask "which of its core abilities remain intact?" Our framework probes four deployment-facing dimensions: procedural reasoning and constraint execution, knowledge grounding and truthfulness, social cognition and affective understanding, and persona fidelity and stylistic consistency. Across more than 35 compressed checkpoints from six open-weight LLMs-covering quantization and pruning-we find that compression redistributes ability profiles. Coarse, local, and surface-level signals often persist, while fine-grained, long-tail, and context-dependent abilities become fragile. Quantization consistently outperforms pruning in retention, but the practical meaning of that gap depends entirely on which ability an application requires. Based on our findings, deployers can choose compression methods by the ability profile their application needs, not just by a leaderboard average. And algorithm designers gain diagnostic signals to build compression methods that preserve what matters. We believe that ACE-Compression turns compression from a black-box accuracy drop into an interpretable intervention-one we can measure, understand, and ultimately improve.

★ DeclineCraft: A RAG Abstention Benchmark for Jointly Measuring Combined Retrieval and Parametric Knowledge

Atharva Thakur, Abulhair Saparov, Bruno Ribeiro — Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 81%, Agents 30%

Reliable retrieval-augmented generation (RAG) requires recognizing when neither retrieved documents nor parametric knowledge supports an answer. We introduce DeclineCraft, a dynamic abstention benchmark whose unanswerable questions target Minecraft feature additions after a corpus cut-off, with audits of possible parametric exposure. Later cut-offs support evaluation of future models. Across 27 LLMs, we measure Correct Abstention Rate (CAR; how often models acknowledge insufficient knowledge on unanswerable questions), Answerable Attempt Rate and Answer Accuracy. We find that (1) single-shot retrieval significantly raises CAR in 38 of 40 configurations (model and thinking setting) versus no retrieval. (2) Iterative RAG, in which the model self-assesses its context and can request up to two further retrieval rounds, significantly lowers CAR versus single-shot RAG in 30 of 40 configurations (mean 46.4%), with little change in Answer Accuracy (68.5% to 69.3%). Adding query rewriting and reranking to hybrid (dense & sparse) retrieval lowers CAR for all 12 models tested. (3) Larger models answer more accurately (Spearman $\rho = +0.60$ with total parameters), but neither model size nor release date shows a significant association with CAR. (4) Without retrieval, enabling thinking on the same weights lowers CAR significantly for 8 of 13 models (by 14% points on average), and the responses shift from acknowledged uncertainty to premise denial.

★ DIVE: Benchmarking Full-Duplex Interaction with Visual Engagement in Continuous Streams

Yuhang Chen, Guangyue Liu, Shilong Liu — Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Recent advances in multimodal large language models (MLLMs) have substantially improved spoken interaction and video understanding. Existing online and streaming benchmarks primarily assess causal multimodal perception through timestamped question answering, whereas real-world interaction requires models to continuously identify task-relevant multimodal cues and decide whether, when, and how to respond. This leaves open how well current models can sustain real-time interaction throughout an evolving multimodal stream. To address this gap, we introduce DIVE, a benchmark for evaluating full-duplex interaction with visual engagement under continuous and realistic multimodal streams. It organizes six tasks into three complementary interaction scenarios. (1) Proactive Intervention evaluates whether models can identify emerging visual or semantic cues and intervene at the appropriate moment. (2) Continuous Engagement evaluates sustained participation through continuous visual tracking and visually grounded query interaction. (3) Interaction Coordination evaluates how models coordinate listening and speaking under overlapping and multi-party interactions. DIVE contains 525 audio/video samples with fine-grained human annotations and precise temporal metadata. Its evaluation protocol separates response quality, participation decisions, and response timing. We evaluate representative real-time models using response quality, window response rates, and reaction latency. Experiments show that leading real-time models achieve sub-second average response latency, yet their average window response rates remain at or below 53.77%. Response quality varies substantially across models and tasks, with adapting ongoing responses to newly arriving information remaining particularly challenging. We hope DIVE facilitates progress toward more natural and capable real-time multimodal interactive systems.

★ OptionWorld: A Benchmark for Diagnosing Failures of LLMs as Long-Term Personal Assistants

Ruoyu Zhang, Zhaocen Liu, Weiyi Tian, Melissa Feng, Shilong Liu — Wed, 21 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 55%

Long-term assistants must follow unfinished tasks across days as information and circumstances change. Finding information or starting an action, however, does not ensure that the task gets completed. We introduce OptionWorld, a benchmark that traces information acquisition, judgments, actions, and outcomes in simulated daily-life tasks to identify where current LLMs fall short as long-term personal assistants. In runs with archived decision inputs, the key facts remain in the model's last input before the deadline in 35.5% of failed instances. Failure analysis identifies the three most frequent problems: needed information is not obtained or combined, available opportunities to act go unused, and agents impose restrictions that the task does not require. Three follow-up experiments show that supplying missing facts restores task success in some cases; action reminders can restart work while subsequent steps still miss the deadline; and correcting a supplied note that wrongly forbids contact restores permitted inquiry. These findings locate shortcomings in maintaining task state: connecting facts to unfinished work, following through before deadlines, and separating actual restrictions from the agent's assumptions.

★ Omni-Woodpecker: Hallucination Correction via Evidence-Prior Rebalancing for Omni LLMs

Zijian Liu, Chaoyou Fu, Yan Yang, Caifeng Shan, Ran He — Wed, 21 Oct 2026 00:00:00 +0000

97% · Hallucination 97%, New model 61%

Recent Omni LLMs (OLLMs) have demonstrated remarkable capabilities in joint audio-visual reasoning, yet cross-modal hallucinations remain a major challenge. Existing mitigation methods mainly operate at the output level, overlooking the entanglement of grounding evidence and learned priors within fused representations. To address this issue, we propose Omni-Woodpecker (OWP), a training-free framework that shifts hallucination mitigation from output-level manipulation to conflict-conditioned representation rebalancing. Specifically, OWP first identifies cross-modal conflicts from complementary modality-specific views. It then employs orthogonal decomposition to separate grounding evidence from competing priors within the hidden states. Finally, OWP adaptively rebalances fused representations by suppressing misleading priors and restoring ignored evidence before decoding. Extensive experiments demonstrate the effectiveness and strong cross-architecture generalization of OWP. Using Qwen2.5-Omni-7B as the backbone, OWP improves performance by 12.9 and 8.6 points on CMM and AVHBench, respectively, while consistently achieving gains of 0.67-2.25 points across four general OLLM reasoning benchmarks. Furthermore, OWP generalizes effectively to VideoLLaMA2-AV, improving performance by 4.6 and 3.9 points on the same benchmarks. Code is publicly available at https://github.com/ZiiJianLiu/Omni-Woodpecker

★ Evolving Self-tuning Benchmarks for Long-Horizon Tool Use

Hyunji Alex Nam, Michael Dennis, Natasha Jaques, Minsuk Chang — Wed, 21 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 96%

For AI assistants to move beyond simple chat interfaces and reliably assist users with applications like Gmail, Venmo, and Slack, we need to improve their tool-use capabilities. However, existing tool use benchmarks are still limited and difficult to scale because they rely on human supervision or are too domain-specific. To make data generation scalable, we propose, TURTLE (**T**ool **U**se **R**elated **T**asks with **L**ong-Horizon **E**volution), an Unsupervised Environment Design (UED) approach to evolve existing tasks using a proposer, a solver, and a feedback generator large language model agents. The proposer is rewarded for difficult yet solvable tasks based on the solver's empirical performance, and the feedback generator summarizes the solver's execution traces for the proposer to exploit. TURTLE produces evolving, self-tuning test suites adaptable to different environment requirements, from supporting user simulators and dense document retrieval to interacting across multiple apps. Empirical results on AppWorld, $\tau$-Bench, and AutomationBench using Gemma 4 31B-it show that the evolved tasks target the desired difficulty range (e.g., a 165\% increase in the percentage of hard tasks on AppWorld), increase task diversity (in terms of reasoning and simultaneous tool interactions), and elicit a broader range of failure modes than the original tasks.

★ ProMemAgent: Long-Horizon Proactive Assistance with Context-Aware Prospective Memory from Sensory Perceptions

Lilin Xu, Linyang He, Kaiyuan Hou, Yuang Fan, Minghui Zhao, Xuhai Xu — Wed, 21 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Large language model (LLM) agents are moving beyond reactive responses toward proactive assistance that anticipates user needs without explicit requests. However, existing proactive agents primarily address immediate needs, while many everyday situations involve delayed intentions that can only be acted on hours or days later. Humans handle such intentions through prospective memory, the ability to remember to act on an intention when the right moment arises. In this paper, we introduce ProMemAgent, the first proactive agent to incorporate prospective memory over everyday sensory contexts for long-horizon proactive assistance. ProMemAgent organizes past sensory perceptions (e.g., egocentric video and audio) into context-aware memory, generating prospective cues and linking earlier events with subsequent developments for future retrieval. Given the current context, it retrieves relevant past events and performs memory-augmented proactive reasoning to determine whether assistance is needed. We further introduce CPMBench, the first benchmark for evaluating long-horizon proactive assistance over sensory contexts, comprising 1,000 instances across eight everyday categories. Extensive experiments show that ProMemAgent achieves the best memory retrieval performance and consistently outperforms all baselines on long-horizon proactive assistance.

★ TeCoP: Accelerating Diffusion Large Language Model Decoding via Tentative Context Pipelining

Jui-Chien Tsou, Hsuan-Chi Liu, Wei-Chin Wang, Yu-Syuan Xu, Chun-Yi Lee — Wed, 21 Oct 2026 00:00:00 +0000

97% · Efficiency 97%

Diffusion large language models (dLLMs) predict multiple token positions per forward pass, yet block-wise decoders serialize generation and repeatedly recompute stable predictions. We attribute this inefficiency to coupling a prediction's readiness for irreversible commitment with its utility as context. We introduce Tentative Context Pipelining (TeCoP), a training-free decoder that decouples these roles through staggered block-parallel pipelining and draft-and-verify. TeCoP reuses stable but uncommitted predictions as provisional cross-block context, with commitment governed by confidence and stability checks. Across four LLaDA and Dream dLLMs on GSM8K, Minerva MATH, and MBPP, TeCoP reduces number of forward evaluations by up to 87% and achieves up to 7.50-fold throughput without KV caching and 11.77-fold with caching over the uncached original decoder. It outperforms Fast-dLLM and DSB in throughput across all 12 model-benchmark settings, with and without caching, while keeping average cached accuracy within one percentage point of the original decoder.

★ RSI Control: Can AI Models Perform Covert Sabotage in the Self-Improvement Process?

Kaiwen Zhou, Udari Madhushani Sehwag — Wed, 21 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Recursive self-improvement, in which frontier AI models work on training the next generation, is quickly becoming a reality. However, the safety of this process has received far less attention than it should. A misaligned model training its successor can shape what that successor knows, how it behaves, and what it wants, and the harm carries forward into the next generation. We study this risk through the lens of AI control. We introduce RSI Control Arena, a testbed in which a frontier model, acting as an autonomous R\&D agent, is given a benign main task from a stage of a successor model's training pipeline together with a covert side task to install a misaligned property in the successor. The testbed contains 16 side tasks in three classes, forming 22 task pairs with main tasks that span the pipeline from data curation to reinforcement learning. We evaluate six frontier models as agents against LLM monitors that read the transcript of the run and the agent's reasoning. We observe that every model completes a substantial share of the side tasks without degrading the main-task performance, including side tasks that make the successor itself a better saboteur. Under minimal instructions to stay covert, models already produce successful attacks that monitors fail to detect, and providing a short list of evasion tactics further reduces the detection rate. Monitors with access to agent's reasoning detects most attacks, but not all. Our results indicate that recursive self-improvement raises significant safety concerns, and that strong safety effort should be in place before models are trusted with greater responsibility of training their successors.

★ CollabSciReM: Aligning Language Models for Scientific Conversations via Science-Aware Reward Decomposition

Sasi Kiran Gaddipati, Gollam Rabby, S\"oren Auer — Wed, 21 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Hallucination 94%, New model 90%

Large Language Models (LLMs) are increasingly used as research collaborators for hypothesis refinement, methodological planning, and technical analysis across multi-turn interactions. In this setting, a useful assistant must be accurate, methodologically justified, logically consistent, and token-efficient. Current alignment approaches rarely achieve all of these together. Instruction-tuned models gain conversational quality through verbosity, and we show that science-agnostic general collaborative rewards compress responses at the expense of scientific rigor. We propose CollabSciReM, a goal-conditioned reward framework that decomposes the session-level reward into conversational quality, scientific rigor, and structural efficiency. We also introduce SciQuality, a pairwise evaluation of multi-turn research conversations grounded in 300 held-out scientific papers and scored on six dimensions: methodological soundness, justification, logical consistency, factual correctness, novelty, and responsiveness. Fine-tuned on 8.62K synthetic research conversations, CollabSciReM is preferred on SciQuality over a science-agnostic reward baseline trained on the same data for all four Llama and Qwen models, including a 65.3% win rate at matched response length. For the 7B and 8B models, CollabSciReM outperforms the science-agnostic reward baseline on the primary metric in 8 of 10 single- and multi-turn benchmark settings. In a blind study, domain experts rated CollabSciReM significantly higher on novelty than the general-reward variant and reported fewer hallucinated responses than both the general-reward variant and the base model.

★ Reading Too Much Into It? Interpretive Preferences and Commitment in LLM Mentalisation under Ambiguity

Stephanie Fong, Yiwen Jiang, Qingyang Xu, Zimu Wang, Yaling Shen, Xiangyu Zhao — Wed, 21 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Large Language Models (LLMs) are increasingly used for interpersonal advice, potentially shaping how users interpret ambiguous social encounters. Existing Theory of Mind evaluations typically rely on predefined answers, while uncertainty research overlooks LLMs' preferred explanations and confidence under psychological distress. To address these gaps, we introduce MentalLoom, a novel evaluation of LLM mentalisation comprising 604 ambiguous social episodes, matched depression and social anxiety disclosures, and four-turn dialogues with unchanged event evidence. Across eight LLMs, disclosures initially elicited more negative explanations overall but more positive narrator-directed interpretations. Scripted repeated questioning showed greater emotional support, stronger positive narrator-directed endorsement, and higher certainty without new event evidence. These findings show that user context can reshape interpretive preferences and commitment while the underlying ambiguity remains unresolved.

★ RACE: RoPE-Aligned Attention Compression Based on Query-Key Activation Energies

Wed, 21 Oct 2026 00:00:00 +0000

97% · Efficiency 97%, New model 59%, Evaluation 37%

Head-dimension compression can reduce the memory required by the key-value (KV) cache, but existing methods often reconstruct full-dimensional representations during inference, adding significant decoding FLOPs overhead. We formalize the design considerations for converting pretrained attention into a natively reduced-dimensional form that reduces both KV-cache storage and attention computation using a calibration set. Specifically, we consider reductions that (1) apply a shared projection to queries and keys so that they remain in a common reduced feature space, (2) derive this projection jointly from query and key activations to better preserve their interaction, and (3) are compatible with the pretrained RoPE transformation. To obtain a reconstruction-free realization of this reduction, we restrict it to coordinate selection and theoretically show that RoPE compatibility then holds if and only if complete rotary pairs are retained. Building on this characterization, we introduce RACE, a post-training method that selects rotary pairs using their joint query-key activation energy. We theoretically show that this criterion minimizes a separable, calibration-dependent upper bound on post-RoPE query-key product approximation error. The resulting reductions are folded into the model weights and require neither runtime reconstruction nor specialized kernels. Across three model families and different compression levels, RACE outperforms structured pruning and remains competitive with reconstruction-based methods on language-modeling, zero-shot, and long-context tasks. At 50% KV-cache retention and 256K context, RACE reduces total inference FLOPs by 47.8%, yielding 1.79x higher throughput, 43.2% lower time-to-first-token (TTFT), and 28.1% lower allocated GPU memory in comparison to an uncompressed model.

★ When Copying Is Hard: Copy-Constrained Decoding for Exact Span Reproduction

Jinghui Zhang, Lang Gao, Zongfang Liu, Ruihong Zeng, Zirui Song, Rui Yan — Wed, 21 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Efficiency 85%, Hallucination 49%

Precisely reproducing a target span verbatim within a long context is a fundamental capability test for LLMs: even when the target string is present in context, models often still drift within the span or fail to stop correctly, revealing limitations in source grounding and boundary control. This capability has broad practical relevance. For instance, code generation often requires embedding sensitive strings such as API keys exactly as they appear in context. In this work, we construct a benchmark to test exact span reproduction ability across contracts, scientific literature, code, and random sequences, and find that current models still make surprisingly frequent errors. To address this, we propose CopyGen, a lightweight copy-constrained plugin for frozen LLMs that turns copying from a byproduct of next-token prediction into a controlled process. Our key idea is to decompose copying into three distinct decisions: when to copy, how far to copy, and when to stop. Concretely, CopyGen first determines whether to enter copy mode, then predicts a tentative copy length, and finally verifies the generated span token by token. During verification, we adaptively adjust the logits of copy continuation tokens to incentivize or suppress continuation, and emit only the longest valid prefix. Across three LLM backbones and four benchmark tasks, CopyGen improves average exact match by 35% relative without model finetuning, and integrates seamlessly with SFT to further improve it by 4%, while substantially accelerating generation. These results suggest that reliable exact copying is not something LLMs obtain for free, but a decoding-time control problem that requires explicit modeling. Code is available at https://github.com/mbzuai-nlp/CopyGen.

★ BitMTP: When Multi-Token Prediction Meets Low-Bit Large Language Models

Ning Zhang, Shihao Wang, Jinrui Zhang, Chaodong Xiao, Lei Zhang — Wed, 21 Oct 2026 00:00:00 +0000

97% · Efficiency 97%, New model 83%, Evaluation 43%

For efficient deployment of large language models (LLMs), methods of ultra-low-bit backbones reduce weight storage, memory traffic, and linear-layer execution cost, while multi-token prediction (MTP) methods reduce sequential decoding steps by predicting multiple future tokens in parallel. Although complementary, the direct combination of these methods cannot accelerate decoding at comparable quality: masked drafting lacks intermediate token input supplied in next-token training, while wider computation and causal key-value (KV) updates can offset the savings from fewer steps. To address these challenges, we propose BitMTP, an algorithm-system co-design framework for efficient MTP on ultra-low-bit LLM backbones. A joint next-token prediction and MTP objective equips a single model with both autoregressive (AR) generation and block-causal drafting. Confidence-only decoding, speculative-only decoding, and hybrid decoding provide different quality-throughput operating points without using separate draft and verifier models. Accepted-prefix alignment keeps the committed prefix, materialized causal KV boundary, and CUDA Graph replay state consistent for continued generation. BitMTP uses this contract to select width-specialized low-bit execution and reuse captured graphs over shared KV storage. On NVIDIA H20, BitMTP with hybrid decoding achieves five-task mean speedups of 2.2x on BitNet b1.58-2B and 2.1-2.4x on 1.7B-4B Bonsai models at comparable task quality. On RTX 3090 and Jetson Orin, it further achieves up to 2.0x in five-task mean throughput and 3.3x on MBPP+, respectively. These results demonstrate that BitMTP accelerates ultra-low-bit LLM generation across backbones, model sizes, and GPU types. The source code is released at https://github.com/PolyU-VCLab/BitMTP

★ Large Flow Language Models

Wed, 21 Oct 2026 00:00:00 +0000

97% · New model 97%, Efficiency 96%, Evaluation 41%

Continuous flow language models offer a path to parallel generation with tools for control and acceleration unavailable to discrete diffusion, yet effective techniques for building them at the billion-parameter scale remain less established. We present Lafa (Large flow language model), a fast, high-performing flow language model scaled to 8B parameters in a data-efficient manner. Lafa is a blockwise flow model with causal attention on the prefix and bidirectional, asynchronous flow within each block. This design enables efficient training by adapting pretrained autoregressive (AR) models and fast inference via prefix caching and blockwise parallel decoding. We train Lafa at 0.5B, 7B, and 8B scales with fewer than 30B tokens, more efficient than previous diffusion models, and evaluate it on instruction following, reasoning, and coding, where the learned flow matches or surpasses the quality of AR and can be distilled into flow maps for one- and few-step generation. Notably, flow training preserves or even improves pretrained AR capabilities via a multi-token prediction effect, which we leverage in a self-speculative decoding scheme that verifies flow-based drafts autoregressively within a single model. This yields 2.54x the throughput of an AR model at comparable quality, outperforming multi-token prediction and discrete diffusion methods. Our findings show that continuous flows can serve as a scalable foundation for language modeling.

★ What Do RTL Repair Benchmarks Actually Measure?

Yeyin Eva Zhu, Rishabh Ranawat — Wed, 21 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

RTL repair benchmarks should measure whether a model can fix broken RTL given a specification. To this end, we construct and hash-bind RTLRepairDataset, 5,784 (broken, fixed) training pairs; RTLRepairBench, 459 held-out cases with 374 syntactic and 85 behavioral mutations; and an evaluation harness supporting simulation and formal equivalence. We then use this infrastructure to identify two measurement confounds. First, simulation and formal scoring give opposite signs for the apparent effect of providing a specification in one evaluated cohort; reset-aware rescoring attributes almost all of the simulation-formal disagreement to reset and first-cycle comparison semantics. Second, on 274 mined model failures, the specification-only minus specification-conditioned recovery point estimate ranges from -8.5 to +18.3 percentage points across six models. Thus, a conventional repair score need not isolate whether the model used the faulty RTL. We translate these findings into evaluation controls: report initialization, reset, first-cycle, and unknown-value semantics; disclose oracle coverage and undecided cases; and include a specification-only baseline. Together, the resources and audit provide both a benchmark for behavioral repair and a protocol for diagnosing what contributes to its measured score. We release the benchmark, the training set, construction code, sanitized aggregate evidence, and reproduction materials at https://github.com/Evaaaz/RTLRepairBench.

★ Dynamic System Inspired Continuous Large Language Model for AI4Science Tasks

Zhaoxin Fan, Zheng Zhu, Yuqing Ma, Faguo Wu, Jian Yang, Wenjun Wu — Wed, 21 Oct 2026 00:00:00 +0000

96% · New model 96%, Evaluation 33%

Scientific foundation models increasingly operate on continuous phenomena, yet still use discrete residual computation. We ask whether pretrained residual transformations can support additional latent computation without expanding the backbone. To test this possibility, we introduce Continuum-LM, which reuses these transformations as dynamical operators in a continuous-depth language model. Its hidden states evolve through an ordinary differential equation evaluated by higher-order numerical integration, with solver order and step size controlling computation. Across eight scientific tasks, Continuum-LM trains stably and improves mean exact-match accuracy by 7.77 and 10.96 percentage points over discrete counterparts at the 0.6B and 1.7B scales, respectively. At 0.6B, it also outperforms an Ouro-style recurrent control under matched parameter and residual-function evaluation budgets. Together, these results support extensible computational depth through parameter reuse, identifying continuous hidden-state evolution as a practical design dimension for scientific language models beyond parameter scaling.

★ PROVISO: Provenance-Informed Action Selection for Secure LLM Agents

Bo Yan, Qi Zhang, Ziwen Kan, Fnu Suya, Song Wang — Wed, 21 Oct 2026 00:00:00 +0000

96% · Agents 96%, Evaluation 46%

LLM agents act on the external world through tools and third-party skills, so a single executed action can cause irreversible harm. Existing defenses judge a proposed action without tracking whether its content comes from the user or from external sources such as tool outputs, and they judge it in isolation rather than against an available alternative, which forces a trade-off between security and utility. We present PROVISO, a controller that decides which action to commit after the agent proposes an action and before it is executed. PROVISO encodes trusted observations, external observations, the proposed action, and an alternative into provenance-typed memory, relates them through role-specific queries, and commits the action with the highest calibrated value learned from outcomes. On four agent security tasks spanning the agent lifecycle, no baseline exceeds PROVISO in both security and utility. PROVISO improves F1 over the strongest baseline by 13.84 points on SkillsBench and by 7.07 points on ToolSafe, and it reaches 88-99 F1 on three additional skill detection benchmarks.

★ SITO: Sifting a Commercial Text-Safety API for Missed Harm and False Flags in English, Ukrainian and Surzhyk

Andrii Tuzhykov — Wed, 21 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

Platforms that host young users screen messages with commercial text-safety APIs, but independent evidence on how often such systems miss harm or flag harmless text is scarce, especially outside English, and the documentation of the one studied here leaves its false-positive rate unquantified. We evaluated one such API, Tuteliq Detect as deployed under the release identifier tuteliq-detect-v5.20261011 (not the separately published v6 model), on six harm categories: authorised push-payment fraud, money-mule recruitment, social engineering, romance scams, gambling harm and bullying. We built SITO (Synthetic Independent Text-safety benchmark with Overflagging tests; sito is Ukrainian for 'sieve'), 2,332 synthetic cases in English, Ukrainian and mixed Ukrainian, with each harmful case matched by hard harmless ones: same-topic messages, minimal-edit contrasts, warnings and retrospective accounts. On English text the system flagged 84.9-94.1% of harmful cases and 0-4.3% of harmless ones per category. It missed third-party reports of ongoing harm far more often than perpetrators' messages (63.6% against 95.1% recall, all languages), when sent without the settings the API offers for reports of harm. Ukrainian and mixed-language input was detected at least as often as English, recall was lower on text from a held-out generator family, so headline recall may be optimistic, and 3.2% of cases received a different decision in repeated runs. A platform should measure false positives on its own traffic and check how bystander reports are handled. The datasets, generation prompts and code are released.

★ MMedAgent-Plus: Scaling Multimodal Medical Agents across Modalities, Tasks and Workflows

Ruoxin Wang, Yuanting Pan, Binxu Li, Ruiyang Ji, Yaqian Chen, Jiayuan Ding — Wed, 21 Oct 2026 00:00:00 +0000

96% · Agents 96%, Hallucination 63%, New model 60%

Medical workflows often require multiple interdependent analyses, but existing specialist and generalist AI models are not designed to coordinate them. We introduce MMedAgent-Plus, a multimodal medical tool-use agent that combines a vision-language planner with 18 leading specialist models spanning 12 tasks and 8 modalities. Through a unified tool interface, the planner selects specialists, executes them in sequence, and synthesizes responses grounded in their outputs. The planner is instruction-tuned on 208,928 samples from 32 public medical benchmarks, covering single- and multi-tool interactions across single and multiple rounds. We separately evaluate tool orchestration, response synthesis, and end-to-end execution on held-out samples. MMedAgent-Plus achieves tool-call accuracies of 91.0%, 98.1%, and 80.7% in single-tool multi-round, multi-tool multi-round, and multi-tool single-round settings, respectively. Given correct tool outputs, MMedAgent-Plus achieves scores of 14.4-14.8 out of 15 for faithfulness, completeness, and conciseness across 600 evaluation samples. In end-to-end execution without pre-supplied tool outputs, MMedAgent-Plus correctly invokes the required specialist in 90.6% of 350 evaluation cases spanning seven tasks. Its final-answer accuracy reaches 0.98, matching the dedicated leading specialist and exceeding the strongest baseline (0.28).

★ Target-Aware Curation of Hallucination Stress Tests for Large Language Models

Jiho Lee, Bingquan Zhang, Lin Ting Liang, Soroush Vosoughi, Weicheng Ma — Wed, 21 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Hallucination 96%

Hallucination stress tests can yield similar average scores across language models even when different prompts elicit hallucinations in each model. We develop a procedure that iteratively edits source-grounded seed prompts to expose hallucinations in a target model. An LLM editor revises each prompt in response to the target's answer and a judge's category-specific score and explanation. In the main setting, the target also receives feedback on its previous response. Experiments cover 500 seeds from seven datasets and ten topics, with intrinsic and extrinsic hallucinations evaluated across open-weight and closed-source models. Editing substantially increases judge-rated severity over the seeds. In the baseline configuration, retained responses fall in the judge's strong-hallucination bin in 77.2% of runs targeting extrinsic hallucination and 74.2% targeting intrinsic hallucination. Later editing rounds account for more than half of the mean score increase. Among 219 prompts that pass model-assisted source-validity checks, 89 elicit responses containing unsupported claims. Across five models, 200 curated prompts yield similar average severity scores, although pairwise Spearman correlations between prompt-level scores are only $\rho=0.0387$ to $0.1897$. Curated prompts can be reused across models, but their diagnostic value depends on source validity and the failures they expose in each model.

★ Unsupervised Tool Evolution for Coding Agents in Data-Lake Analytics

Wed, 21 Oct 2026 00:00:00 +0000

96% · Agents 96%, Evaluation 62%, New model 51%

Coding agents can answer analytical questions over a data lake by discovering, cleaning, and joining the relevant data. However, they discard this work after each question and redo it from scratch for the next one. Tool evolution, which automatically builds and refines reusable tools across tasks, is a natural fit for this repetition. Since a ground-truth label for a data-lake question requires an expert to solve it first, we consider the unsupervised setting. We propose TEAL (Tool Evolution for Analytics over data Lakes), which evolves a library of tools and usage skills for a given Worker agent without labels. Two agents propose tools from complementary evidence, an Explorer from the data lake itself and a Critic from Worker trajectories. A proposed change is kept only if the Worker adopts the tool on a rerun without extra steps and any changed answer is supported by trajectory evidence. We also introduce TEAL-Q, which generates synthetic questions for tool evolution when real ones are unavailable. We evaluate on KramaBench and LakeQA with four Worker LLMs under step budgets, which cap the number of steps the Worker may take per question. With a tight budget of 10 or 15 steps, TEAL improves every Worker on both benchmarks, lifting Sonnet-5 at 15 steps by 9.3 points on KramaBench and 21.7 points on LakeQA. A tighter budget also yields better feedback for evolution. For example, Sonnet-5's LakeQA library evolved at 20 steps and deployed unchanged at 30 steps scores 68.3, above both the baseline (59.6) and a library evolved directly at 30 steps (64.2). Additionally, a library evolved on synthetic questions performs comparably to one evolved on real questions. These results show that useful tools can be evolved without labels, and hopefully inspire future work on what to evolve and how to obtain feedback in unsupervised agent evolution.

★ Detecting Errors in Large Language Model Responses: The Incremental Predictive Value of Hidden States

Ningzhi Li, Xiaojun Ma, Xiaofei Wang — Wed, 21 Oct 2026 00:00:00 +0000

96% · Hallucination 96%, Evaluation 69%

Although large language models support a wide range of tasks, hallucinations remain a major obstacle to user trust. When a model generates an answer, both its internal processing and final output vary with the question. Internal states are often expected to offer richer signals for hallucination detection. However, studies using these states typically report their standalone discrimination performance, while output-side comparisons often use simple baselines with weaker modeling capacity. We integrate output-side information available at inference time into a combined baseline and compare detection performance before and after adding hidden states under identical data splits and evaluation rules. Experiments cover open-ended generation and constrained multiple-choice tasks across different models. Adding hidden states improves AUROC by approximately 0.004 and 0.039 on the two tasks, respectively. Hidden states thus provide additional predictive information, but strong standalone detection performance can still yield only limited additional gains.

★ The Show-Up Attack: When Subagents Win Over Their Supervisor

Maosen Zhang, Jianshuo Dong, Yutong Zhang, Jie Zhang, Tianwei Zhang, Han Qiu — Wed, 21 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

Multi-agent systems are increasingly deployed to coordinate multiple large language model (LLM) agents. However, this also creates new attack surfaces: malicious agents may have chances to exploit the coordination. In this paper, we propose and study the \textit{show-up attack}, in which a malicious subagent manipulates the supervisor into involving it in tasks beyond its assigned scope. To evaluate this risk, we propose SUA-Arena, which implements 6 attack strategies and contains 188 test cases. Then, we evaluate 6 frontier LLMs and find that all of them are vulnerable to show-up attacks. For example, GPT-5.6-Sol as a supervisor suffers an ASR of up to 0.93. Mitigation strategies such as prompting and restricted communication fail to reduce the risk consistently. These results show that multi-agent systems need explicit mechanisms for enforcing authority boundaries.

★ Beyond Calibration: Temporal LLM Confidence Assessment Across Languages and Local Settings

Sahil Kale — Wed, 21 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Hallucination 71%

Large language models are expected to estimate their own uncertainty, with confidence signals used to guide retrieval, adaptive compute allocation, routing, and selective abstention. We investigate whether confidence signals measured at different temporal stages of problem solving encode distinct forms of epistemic information and remain reliable beyond English-centric settings. Inspired by classical meta-memory theory, we distinguish anticipatory pre-solve confidence (Feeling-of-Knowing, FOK) from retrospective post-solve confidence (Judgement-of-Learning, JOL). We evaluate these signals across two frontier LLMs, multiple reasoning domains, translated Marathi benchmarks, and a newly constructed benchmark of genuinely local Marathi factual questions. Across settings, the cheaper pre-solve signal remains weakly calibrated, while post-solve confidence generally provides stronger epistemic alignment despite its additional inference cost. Confidence behaviour transfers surprisingly well under direct language translation, with calibration trends largely preserved between English and Marathi, but degrades substantially on local Marathi knowledge, suggesting that linguistic transfer alone is insufficient for region-specific information needs. Finally, hallucinated responses maintain persistently elevated confidence after answer generation, highlighting important limitations of current LLM self-assessment mechanisms.

★ ProVe: Proposer-Verifier Decoding

Ziqi Jin, Weiwen Xu, Jiaxi Li, Yujie Zhong, Qi She, Aixin Sun — Wed, 21 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, New model 74%, Evaluation 42%

Speculative decoding accelerates autoregressive (AR) language models by letting a small model draft tokens serially and a large model verify them in parallel. We reverse these roles. Starting from a standard AR checkpoint, we fine-tune the large model into a proposer that predicts a block of future tokens in a single forward pass, then use one additional forward pass to verify the proposed block. This gives ProVe, Proposer-Verifier decoding. The verifier can be a smaller AR model or the proposer itself; the latter gives ProVe-lossless, which reproduces the proposer's greedy AR decoding. Block proposals already provide substantial room for acceleration: with a block size of 16, the proposed block has an average AR-matching prefix of 6.6 tokens on GSM8K. Across six benchmarks with a 32B proposer, ProVe-lossless reaches 2.5x average AR throughput without training a separate draft head. With a small 3B verifier, ProVe reaches 2.7x average throughput at 2.6 accuracy points below AR on the six-task average, providing a controllable tradeoff between speed and accuracy.

★ Polarity Blindness in LLM Safety: A Systematic Study of Negation, Antonym Substitution, and Adversarial Framing

Diego Miguel Lozano, Erik Derner, Juan Antonio P\'erez-Ortiz, Nuria Oliver — Wed, 21 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

Negation remains a persistent challenge for language models across NLP tasks and modalities. We show that this weakness can undermine LLM safety guardrails through a structural vulnerability we term polarity blindness: the failure to recognize harmful intent after polarity-altering linguistic transformations. We study two forms of polarity inversion: grammatical negation and antonym replacement, and use synonym replacement as a polarity-preserving control. We further test whether an adversarial textual wrapper amplifies these failures. Using a 260-prompt benchmark across 10 state-of-the-art LLMs and 3 languages (English, Spanish, Czech), stratified by harmfulness potential and evaluated on a validated 0-5 harm scale, we find that polarity blindness affects most evaluated models, with mean harm scores reaching 4.12 for the most vulnerable model. The effect persists across typologically differing negation systems, suggesting that polarity blindness reflects a broader limitation in semantic robustness rather than a language- or grammar-specific failure.

★ Lost in Recovery: Demystifying Recovery Behavior of Computer-Use Agents via Goal-Oriented Action Replay

Wed, 21 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Agents 90%

★ Cache Engineering for Agents

Tianyu Fu, Ruoxi Wang, Zihan Min, Zhongqi Ren, Zhihao Li, Tengxuan Liu — Wed, 21 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, Agents 70%, Evaluation 37%

Large Language Model (LLM) agents can handle increasingly complex tasks, navigating long trajectories of observations and actions. As these trajectories grow, key-value (KV) cache I/O and storage become the dominant cost. Trajectories can also exceed the LLM's maximum context length, requiring control over active cache growth. Prior work uses context engineering to reduce cache size by trimming or summarizing history. However, modifying the trajectory prefix invalidates subsequent cached states and triggers costly re-prefilling of retained context. Task accuracy can also suffer when too little context is retained to limit re-prefill costs. We propose cache engineering as a new scheme. It allows agents to directly control their active KV cache without modifying the trajectory or re-prefilling retained context. Specifically, agents can proactively use drop to exclude selected cache from future attention and repos to adjust retained keys' position encodings. We formalize these cache operations and identify history dependence as a serving challenge: identical token sequences can yield different cache states under different cache operation sequences, making prefix matching insufficient for correct cache reuse. Thus, we co-design the AKV serving engine. It provides a stateless interface compatible with existing serving APIs, while internally managing stateful cache reuse, eviction, and restoration. Experiments show that a tool-response rolling-drop cache policy improves average benchmark scores by 1.7 points and reduces average cost by 27.7%, compared with conventional threshold-triggered summarization. With AKV, it achieves up to 1.2x the output throughput and a 1.5x speedup in P95 request latency. Recursive self-improvement of cache policies further reduces cost by 7.1% over expert design. Our code is at https://github.com/thu-nics/AKV.

★ Value-Rich, Differently Organized: Measuring Value Expression and Organization in AI-Generated Debates

Maoyuan Li, Haoyuan Li, Zhongsheng Wang, Jiamou Liu — Wed, 21 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

Current evaluation of LLM-generated debate largely focuses on argumentative quality, leaving the value structure of generated arguments underexamined. We introduce a diagnostic metric framework for comparing how values are expressed and organized in human-reference and AI-generated debates. The framework separates value expression, the range of recognizable values a debate invokes, from value organization, how those values are distributed and sustained across turns. Grounded in Schwartz's Basic-10 value theory, it represents debates as multi-level value distributions over argumentative units, utterances, stances, and full debates, enabling metrics for value coverage, alignment, diffusion, and cross-turn dynamics. Across multiple generation systems, AI-generated debates are broadly value-expressive, often activating as many or more value dimensions than matched human references, but they organize these values differently. Their utterances are more locally value-diffuse, and source-blind judges distinguish value clarity from value focus. These findings show that debate evaluation should consider not only value presence, but also value organization. Code and supplementary materials are available at https://github.com/Maoyuan-li/Value-Expression-and-Organization.

★ Adaptive Multi-Token Prediction: How far should we go?

Emile Anand, Mohammed Rehan Parwani, Xinyuan Cao — Wed, 21 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, Evaluation 38%

Multi-token prediction (MTP) can accelerate language-model decoding by drafting several tokens and verifying them in a single target-model pass. However, the best speculation horizon changes with context: longer drafts offer greater amortization when reliable, but can waste compute on suffixes discarded after a rejection. Unlike autoregressive drafting, where a stopping rule can decide token by token, MTP must fix the horizon before drafting, as all heads fire in the same forward pass. We therefore formulate horizon selection as an online control problem, using verification feedback to learn how far the draft can be trusted in each context, and introduce Adaptive Multi-Token Prediction (AMTP): a lightweight controller that requires no retraining and leaves the exact verifier unchanged. To guide horizon selection, we characterize the expected number of tokens committed at each horizon via acceptance survival, the probability that a draft prefix is fully accepted. Before each round, AMTP predicts how far the draft is likely to survive and picks the horizon with the highest expected tokens per second, learning freely from every verification pass. Theoretically, we show that any online learner's prediction error translates into throughput regret against the best horizon for each context, and in turn into a bound on wall-clock completion time. With online gradient descent, this regret is sublinear. Across five draft-target pairings built on Llama-3.1-8B-Instruct with Medusa heads, Meta's MTP models, and Qwen3-4B-Inst-2507-MTP, evaluated on AIME 2025 and GSM8K, AMTP decodes up to 4.1x faster than standard decoding and matches or outperforms the fixed window in 8 of 10 settings, with throughput gains of up to 10%.

★ LightCRS: Lightweight Multi-Domain Conversational Recommendation System

He Ma, Lan You, Qiyang Wu, Chen Liu — Wed, 21 Oct 2026 00:00:00 +0000

95% · Evaluation 95%, Efficiency 93%, New model 54%

Conversational recommender systems have recently adopted vision-language models to match user dialogue intents with multimodal product information. However, extending these methods to many product domains faces a dilemma: deploying a separately adapted VLM per domain scales cost linearly with the number of domains and requires routing each dialogue to the right model, while naively merging all domains into a single VLM degrades accuracy because product characteristics vary widely across categories. We propose LightCRS, a lightweight, query-conditioned framework for unified multi-domain multimodal conversational recommendation. LightCRS precomputes multimodal product representations once offline with frozen encoders. Online, a compact 5.3M-parameter module fuses this evidence for each dialogue via query-conditioned cross-attention, whose parameters are shared across all domains, and compares candidates jointly with an inter-candidate Transformer. To enable rigorous evaluation, we construct MM-R2A-25, a 25-domain benchmark pairing natural Reddit recommendation-seeking conversations with Amazon product listings. LightCRS achieves state-of-the-art average HR@1, NDCG@5, and MRR, improving HR@1 by 35.4\% over the strongest baseline while reducing per-query latency by 266 times. Joint multi-domain training itself improves HR@1 by 15.2\% over separately trained single-domain models.

★ Diagnostic Analysis for Large Language Model Reasoning

Xinyu Liu, Dongqi Fu, Jun Wu — Wed, 21 Oct 2026 00:00:00 +0000

95% · Evaluation 95%, Hallucination 51%

Reasoning structures (e.g., chains and graphs) are increasingly used for probabilistic inference, factuality verification, and uncertainty estimation, but their structural reliability is rarely evaluated directly. We test whether this reliability holds through a proposed two-stage diagnostic evaluation in open-ended multi-hop QA. Stage 1 tests different capabilities required to construct reasoning graphs, showing that graph reliability varies with topology, prompt design, and model family. Stage 2 uses uncertainty estimation as a downstream test case. Generated-edge experiments show that graph-based computation is sensitive to upstream graph errors, with performance dropping relative to the gold-graph setting. Gold-graph experiments further show that graph structure provides little benefit when used only as prompt context, but improves the separation of correct from incorrect answers by confidence score and enables node-level diagnostics.

★ Global Security Blind Spots: Translation as a Safety Bypass in Low-Resource Languages

Wed, 21 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

Large language models are deployed globally, yet safety alignment and evaluation remain concentrated in English: harmful requests refused in English can elicit unsafe completions when translated into low-resource languages. We evaluate ten widely deployed LLMs across 79 languages (English plus 78 low-resource) under a translation-bypass threat model, using a metric conditioned on English safety, which separates translation bypass from unsafety already present in English. English safety does not transfer reliably: conditioned on safe English completions, mean bypass rates range from 21.9% to 44.4%. Bypass risk concentrates in a subset of language families, a pattern broadly shared across models. The shared ordering is confounded with translation degeneracy, but its variation across models is significant and survives every control we apply. We also show that per-model bypass rates do not identify a stable ranking: nine of ten models change rank across three conditioning schemes that estimate the same quantity, with no change to any underlying response.

★ DStill: Trajectory Consistency Distillation for Accelerated One-Step Drafting in Block Diffusion Speculative Decoding

Wed, 21 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, New model 86%

Block-diffusion speculative decoding accelerates large language model inference by drafting multiple tokens in parallel. However, one-step drafting in discrete diffusion is limited by token factorization error: each token is predicted from the same masked block, without conditioning on the tokens selected at other positions in that step. This can produce inconsistent drafts and reduce acceptance length. Iterative refinement mitigates this limitation by conditioning later predictions on previously committed tokens, though additional passes increase drafting cost. To address this trade-off between draft fidelity and inference latency, we introduce DStill, a two-stage framework that distills a target-conditioned drafter's own multi-step refinement into one-step drafting. We first train the drafter to complete partially revealed blocks, then use its refinement trajectories to train a student to predict the final completion in one pass. Using a curriculum that gradually transitions from partial to fully masked inputs, we successfully transfer the multi-step refinement capabilities into one step. Across reasoning, math, code benchmarks on three model families DStill improves acceptance length by up to 28% and speedup by up to 25% over one-step DFlash, achieving up to 6.45x speedup over auto-regressive decoding.

★ When LLMs Manipulate: Measuring Behavioral, Social, and Temporal Effects

Wed, 21 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

Large language models (LLMs) can influence user trust and decision-making through subtle linguistic cues, yet the behavioral effects of manipulative outputs remain poorly quantified. We introduce a behavioral computing framework for measuring \textit{behavioral vulnerability}, defined as users' susceptibility to manipulative AI responses. Building on DarkPatterns-LLM benchmark, we generate controlled manipulative stimuli using Manipulation Gradient Descent (MGD), targeting biases associated with authority, urgency, and social proof. We analyze participant responses using the Multi-Stakeholder Impact Assessment Network (MSIAN) to model cross-user influence and the Temporal Harm Propagation Model (THPM) to characterize the persistence of behavioral effects over time. In a study involving 120 participants, manipulative framing produced significant changes in trust calibration and compliance ($p < 0.01$), with effects also propagating across social contexts. These findings highlight the need to extend LLM safety evaluation beyond model outputs to measurable behavioral consequences and provide a computational framework for quantifying individual, social, and temporal risks arising from manipulative AI interactions.

★ Self-Decomposing Diffusion for Code Generation

Lize Shao, Zichen Xie, Zhizhen Chen, Zhe Zeng, Wenxi Wang — Wed, 21 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, New model 44%, Evaluation 38%

Diffusion language models (DLMs) are emerging as a promising paradigm for code generation, offering the potential for fast, parallel decoding. However, generating complex programs remains challenging, as their interdependent components must maintain consistent interfaces, data flows, and shared assumptions. Aggressive parallel decoding can compromise these dependencies and undermine program correctness, while conservative decoding sacrifices efficiency. To address this trade-off, we propose planning program decomposition before implementation to coordinate interdependent units during parallel generation. We introduce Self-Decomposing Diffusion for Code (SDDC), a novel framework that decomposes complex programs and implements their units in parallel, entirely through diffusion. SDDC first generates a contract specifying the program's decomposition and inter-unit dependencies, then uses it to guide coordinated parallel implementation. Across five benchmarks and two 7B diffusion backbones, SDDC substantially improves the accuracy--latency trade-off, raising pass@1 by up to 31.1 percentage points while decoding 7.5--13.7$\times$ faster than the base models, and outperforming a fine-tuning baseline by up to 23.5 points with lower latency.

★ SCODEL: Shared Canonical Ontology Definitions for Food Entity Linking and Ontology-Grounded Recipe Knowledge Graphs

Wed, 21 Oct 2026 00:00:00 +0000

95% · Evaluation 95%, New model 60%

Identifying recipe entities and linking them to ontology concepts is fundamental for semantic recipe understanding. However, food mentions in recipes are often ambiguous, context-dependent, and lexically diverse, making ontology entity linking (EL) challenging. We propose SCODEL, a streamlined, model-agnostic approach based on a central idea: contextual food mentions and ontology concepts that refer to the same entity should learn the same canonical ontology-oriented definition. SCODEL uses these shared definitions as a semantic normalisation layer, providing complementary semantic information to existing EL models without requiring a new linking architecture. We further introduce ReciLinkD, the first benchmark for contextual food ontology entity linking in recipes. Experiments show that SCODEL consistently improves three modular EL models (ReFinED, GSM-EL, and BLINK): SCODEL x BLINK achieves the highest Hit@1 on seen concepts (88.33%), while SCODEL x GSM-EL achieves the highest Hit@1 on unseen concepts (83.15%), outperforming the strongest large language model (LLM) based EL baseline in each setting. Experiments on a biomedical ontology benchmark further demonstrate that SCODEL generalises beyond the food domain. Finally, we use SCODEL to ground food mentions at scale and construct an ontology-grounded recipe knowledge graph, demonstrating its utility for ontology-aware recipe retrieval and corpus-level food analysis.

★ GRADUS: Discovering and Recovering Agent Lessons from the Environment's Latent Syllabus

Qing Ping, Wei Yang, Panpan Xu, Johannes Kirmayr, Lukas Stappen, Junjie Tang — Wed, 21 Oct 2026 00:00:00 +0000

95% · Agents 95%, New model 90%, Evaluation 35%

Training LLM agents for policy-constrained tool environments, such as an airline backend with booking, refund, and cancellation policies, requires tasks that exercise the environment's rules individually and in combination. Existing synthesis pipelines let a model propose tasks and retain only those it immediately solves, biasing training data toward easier behaviors. We introduce Gradus (Graded Rungs of Ascending Difficulty from an Underlying System), which decouples what a task should teach from the concrete task and solution used to teach it. Without seed tasks, Gradus treats a fixed environment as a latent syllabus, extracting guarded behavioral operators that capture how actions depend on current conditions and how their effects reshape subsequent decisions. It composes these operators into increasingly demanding tasks while using corpus-level feedback to target missing outcome combinations. Crucially, Gradus jointly calibrates tasks and solutions against the environment: execution feedback identifies which side to refine while preserving the intended behavioral requirements. Verification thus becomes a recovery signal rather than a terminal filter, coupling coverage-guided exploration with the preservation of difficult examples. From off-the-shelf backbones, the resulting curriculum lifts a 30B model from 59% to 89% on $\tau^2$-Bench and from 47% to 83% on CAR-Bench with an Opus 4.5 proposer, and to 76% and 62%, respectively, with an open-weight Qwen3-235B proposer. These results show the value of jointly controlling which behaviors are explored and which survive into verified training data.

★ SAPD: Learning from Agent Experience without Environment Interaction

Juncheng Diao, Haibin Wang, Wei Shi, Yifan Ding, Qingbin Li, Rongxiang Weng — Wed, 21 Oct 2026 00:00:00 +0000

95% · New model 95%, Agents 88%, Efficiency 47%

Scaling the training of Large Language Model (LLM) agents typically requires a large collection of executable environments, whose construction, maintenance, and operation constrain the expansion of training experience. Existing offline trajectories offer an opportunity to reuse interaction experience, yet effectively refining the current policy without revisiting these environments remains challenging: supervised fine-tuning (SFT) can lead to catastrophic forgetting, while online reinforcement learning requires real-time interaction with the environment. To address this challenge, we propose Step-Advantage Policy Distillation (SAPD), a method for learning from offline trajectories using relative step advantages. SAPD decomposes multi-turn trajectories into individual decision contexts for on-policy distillation, allowing the student to generate responses conditioned on recorded histories, and introduces relative step advantage signals to stabilize training. The entire process requires neither new environment observations nor additional multi-turn interactions. Experiments on ALFWorld, WebShop, and Search-based QA show that SAPD improves performance across different teacher configurations, mitigates catastrophic forgetting compared with SFT, and runs at least $4.7\times$ faster per training step than online OPD. We will release our code soon.

★ Building Telecom Language Models with RSOD: Domain Data, Specialist Post-Training, Memory-Conscious Distillation, and Networking Evaluation

Wed, 21 Oct 2026 00:00:00 +0000

95% · New model 95%, Evaluation 87%, Efficiency 53%

Telecom language models must combine broad networking knowledge with specialist behavior across routing, broadband services, access technologies, and operational diagnostics while remaining practical to deploy as a single model. We present an end-to-end framework that builds source-grounded synthetic training data from structured networking taxonomies and expert graphs, materializes 212,559 routed training records across fourteen networking routes, trains six frozen specialist LoRA teachers over one shared continued-pretrained backbone, and consolidates their behavior into one deployable student adapter using Routed Specialist On-Policy Distillation (RSOD). In RSOD, the committed student generates each trajectory, the route-assigned specialist teacher-forces the exact native token sequence, compact teacher hidden states are reconstructed through the shared frozen output head, and only the student LoRA is updated. A controlled reconstruction validation shows mean absolute logit error of 7.26 x 10^-6 with 100% top-1 agreement between native and reconstructed teacher logits. On four RTX 6000 Ada GPUs, the completed 2,000-step run produced 16,000 trajectories with zero observed student-version or token- ID mismatches. Under the primary route-conditioned evaluation over 2,850 held-out questions, the raw Meta-Llama-3.1-8B-Instruct model scores 12.8661/25 micro (13.0172/25 macro), the combined routed six-teacher system scores 21.7641/25 micro, and the consolidated RSOD student, released as Llama-CalixCLM v1, reaches 21.7112/25 micro (21.8006/25 macro) on the same five dimension LLM-judge rubric. Under a separate neutral deployment-like prompt, the same student reaches 20.0028/25 micro (20.0778/25 macro). The consolidated student remains within 0.0529 micro-score points of the combined routed-teacher system while retaining a numerically validated and auditable train-serve execution path.

★ FALCA: Falsification-Aware Reinforcement Learning with Hierarchical Credit Assignment for Long-Horizon Deep Search Agents

Zijun Dou, Kun Luo, Chaofan Li, Shuqi Lu, Zheng Liu — Wed, 21 Oct 2026 00:00:00 +0000

95% · Agents 95%, New model 86%, Hallucination 50%

Long-horizon deep search requires LLM agents to iteratively decompose complex information needs, gather evidence from external sources, and maintain a reliable research state under multiple constraints. A central challenge is that early mistakes can be amplified over the search trajectory: agents may anchor on an incorrect candidate, collect partially supporting evidence, overlook contradictions, and accumulate errors that misguide subsequent reasoning. We propose FALCA, a falsification-aware framework that formulates deep search as a propose-falsify process. FALCA adopts a collaborative Searcher-Falsifier architecture, where the Searcher proposes candidate answers or sub-answers, while independent Falsifiers actively seek counterevidence, check unmet constraints, identify conflicts, and produce concise falsification reports for research-state revision. To strengthen this loop, we introduce falsification-aware training, combining SFT warmup on falsification trajectories with agentic RL based on parallel rollouts and hierarchical credit assignment over research stages, state transitions, and actions. Experiments on five search benchmarks show that FALCA improves long-horizon deep search performance over backbone and harness ablations, while reducing confirmation bias and error accumulation through more effective falsification and research-state revision.

★ Stratified Sparse Attention: Reducing the Sparse-to-Dense Gap at Extreme Sparsity

Wed, 21 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, Evaluation 31%

Large Language Model inference over long contexts is increasingly bottlenecked by the cost of attending to a growing Key-Value (KV) cache. Sparse attention reduces this cost by accessing only a fraction of the cache, but existing methods face a trade-off between deterministic retrieval, which captures high-scoring tokens but discards residual attention mass, and sampling, which improves coverage at the cost of higher variance. Hybrid methods combine top-$k$ retrieval with uniform sampling, yet still exhibit a substantial gap from dense attention at aggressive sparsity. We introduce \emph{Stratified Sparse Attention (SSA)}, a unified framework combining deterministic retrieval with score-guided sampling. SSA deterministically attends to high-scoring tokens while sampling the residual using proxy attention scores and correcting for their inclusion probabilities, thereby recovering attention mass beyond the retrieval set. SSA encompasses top-$k$, top-$p$, and hybrid top-$k$+sampling as special cases. Across RULER, LongBench, LOFT, and AIME, and five models ranging from 3B to 27B parameters, SSA consistently improves over the baselines at matched sparsity. On RULER-HARD with Ministral-3-8B, SSA improves accuracy over the strongest practical baseline from 81.8\% to 84.2\% at 32K and from 63.8\% to 69.3\% at 128K under $50\times$ attention sparsity, recovering 94\% and 85\% of dense-attention performance, respectively. With optimized CUDA kernels, SSA achieves up to $2.96\times$ decode speedup over FlashInfer at 1M context, demonstrating that its accuracy gains translate into practical long-context inference acceleration. We open-source our code at https://github.com/skylight-org/SSA-Paper-Code.

★ Evolving Agent Interfaces

Wed, 21 Oct 2026 00:00:00 +0000

95% · Agents 95%, New model 50%, Evaluation 39%

Computer-use agents are limited by the interfaces humans give them. An agent may type a spreadsheet cell by cell when a single line of code would do, or print all two thousand lines of a file through a shell to find one function. Both the interfaces and the knowledge of when to use each are built by hand, which cannot keep pace with new applications, workflows, and models. We argue that interface design should itself be an agent capability. We introduce Agent S4, an autonomous computer-use scientist that builds Evolving Agent Interfaces from task experience. A meta-agent examines task agents' trajectories on synthetic training tasks, without success labels, and constructs and revises interfaces and usage knowledge in a repository. Population-based evolution then selects which repositories to develop further by their performance on validation tasks the meta-agent never sees, favoring changes that generalize beyond the seen tasks. On OSWorld V2, Agent S4, with Qwen3.8-27B and the evolved repository of interfaces, sets a new state of the art in open-source computer-use agents, reaching 21.3% binary score, 3.8 times the same model with fixed interfaces. This is above both GPT-5.5 (13.0%) and Claude Opus 4.8 (20.6%), at about 4.3% and 1.5% of their respective costs. On ScienceBoard, the evolved interfaces improve the success rate from 41.4% to 58.6% with the same model. These results suggest that useful interfaces can be evolved from experience, making autonomous interface design a source of improvement complementary to advances in models and harnesses. We open-source our code at https://github.com/simular-ai/agent-s.

Do We Need the Largest LLMs for Every Clinical Decision?

Ali H. Lazem, William Teahan — Wed, 21 Oct 2026 00:00:00 +0000

94% · Evaluation 94%, Efficiency 93%

Large clinical language models often require multi-GPU infrastructure, limiting local deployment. We introduce CERTA, a decision-level statistical certificate that determines when a quantized model may substitute for its native BF16 reference and escalates decisions that cannot be certified to it. Across eleven instruction-tuned models, both general-purpose and domain-adapted, from three families (7-72B), we evaluated 2,500 clinical-verification decisions from de-identified patient contexts (2,484 patients), with patient-disjoint calibration and testing. In the higher-power operational analysis, round-to-nearest 8-bit quantization at BF16 storage served 99.8-100% of held-out decisions across all eleven models; formal finite-sample guarantees were assessed separately using the exact split-sample procedure. In the same operational analysis, the class-conditional 4-bit frontier reduced the amortized, BF16-equivalent deployment cost to 25.8% and 27.3% of the BF16 reference for Qwen2.5-72B and HuatuoGPT-o1-72B, respectively, with measured peak memory of 43.0 and 42.7 GB, below the nominal capacity of a single 48 GB accelerator. Deployment on that specific hardware was not directly tested. Under the exact split-sample certificate, which also accounts for configuration selection, these costs were 26.7% and 27.9%. On external disease named-entity benchmarks against human annotations, the frontier-selected NF4 configurations had a mean absolute F1 change of 0.045 on both datasets, with substantial model-specific variation. CERTA converts compression from a fixed precision choice into a risk-controlled deployment policy for this verification task.

Beyond Policy Discrepancy: Capability-Aware Multi-Teacher On-Policy Distillation

Xuanfeng Yang, Le Gui, Jiaming Wang, Xuan Zhang — Wed, 21 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 35%

Multi-teacher on-policy distillation (MOPD) integrates domain specialists into a shared student, but determining how strongly each teacher should supervise remains challenging. Existing discrepancy-based allocation can assign strong supervision even when the student already matches or exceeds the teacher's task performance, since behavioral mismatch does not necessarily imply a remaining capability deficit. We propose Capability-Aware Multi-Teacher On-Policy Distillation (CA-MOPD), which combines specialist merging with capability-guided allocation to account for what the student has already learned. CA-MOPD first merges reinforcement-learning specialists into a strong student initialization, then uses verifier-estimated teacher--student capability gaps to adjust domain-level distillation strength. Supervision decreases as capability gaps close and increases when estimated student capability regresses, while dense token-level guidance remains provided by the teachers. Across six benchmarks spanning mathematics, code generation, and instruction following, CA-MOPD achieves an average score of 32.22, outperforming naive MOPD, Open-MOPD, and the merged initialization by 0.86, 0.83, and 0.74 points, respectively. Controlled comparisons further support the benefits of merged initialization and capability-gap guidance under matched training budgets.

Decomposed Relation Extraction from Aviation Accident Narratives: Evaluating Transformer-Based Safety Knowledge Discovery under Relational Sparsity

Aziida Nanyonga, Mustafa Tahsin Yilmaz, Saad Ezzini, Ugur Turhan, Graham Wild — Wed, 21 Oct 2026 00:00:00 +0000

94% · Evaluation 94%, Hallucination 49%

Relation extraction from aviation accident narratives is challenging because genuine safety relations are sparse among possible concept pairs. This study evaluates a decomposed framework separating relation classification from concept identification. A benchmark of 300 National Transportation Safety Board (NTSB) narratives containing 2,524 unique concept-text nodes was expanded into 21,038 ordered candidate pairs, with only 248 (1.18%) representing RESULTS_IN or CONTRIBUTES_TO relations. A TF-IDF Linear Support Vector Machine (SVM) and weighted RoBERTa classifier were developed using an independent graph-annotated resource with hard-negative sampling and frozen before sealed evaluation. Development Macro-F1 scores of 0.4738 (SVM) and 0.4809 (RoBERTa) declined to 0.2741 and 0.2420 under exhaustive evaluation, with Matthews correlation coefficients of 0.0128 and 0.0197. False positives dominated: the SVM recovered 96 reference relations but generated 7,018 false positives, while RoBERTa recovered 125 but generated 9,824. Post-test diagnostics revealed overlapping RoBERTa confidence distributions and false positives even within valid relation-category directions. Complementary end-to-end SmolLM2 evaluation identified missed and unsupported relations alongside malformed outputs. These findings demonstrate that known concept endpoints do not guarantee reliable relation extraction and that sampled-negative development performance may poorly reflect exhaustive evaluation. Controlling false positives and managing candidate-space sparsity remain essential for aviation safety knowledge extraction.

MLIPS: Multi-Layer Indexer Pool Sharing

Samkit Jain, Sai Kishan Pampana, Shlok Agarwal — Wed, 21 Oct 2026 00:00:00 +0000

94% · Efficiency 94%, Evaluation 31%

Sparse attention limits the keys used by each query, but finding those keys can still require a full-prefix scan at every layer. We introduce \textbf{MLIPS (Multi-Layer Indexer Pool Sharing)}, a training-free method that shares candidate search while retaining each consumer's own scoring and selection. A producer publishes a top-$M$ pool larger than the attention budget $K$; subsequent layers select their own top-$K$ from that pool. Measurements across three sparse-attention model families show substantially higher coverage from broad pools than overlap between final selections. We implement layerwise sharing, persistent decode reuse, and a coarse producer in vLLM for DeepSeek-V4-Flash. Across four long-context suites, three consumer layers per producer give $1.14$--$1.40\times$ speedups and $14.8$--$27.4\%$ lower GPU energy, with observed accuracy changes of $-1.49$ to $+0.29$ percentage points in single runs. More extensive sharing exposes task-dependent quality losses. Sampled prefill traces also reveal recurring selected KV positions across query groups and layers. We characterize this shared working-set structure and use it to motivate Top-$G$, an exploratory bounded pool construction for prefill. Its deployment, task accuracy, and serving benefits remain open.

E8-QAT: Efficient 2-Bit Quantization-Aware Training with Implicit Lattice Codebooks

Xinyu Lei, Hong Huang, Jiebin Zhang, Hongyu Li, Jinlong Song, Decheng Wu — Wed, 21 Oct 2026 00:00:00 +0000

94% · Efficiency 94%, New model 92%, Evaluation 30%

Low-bit quantization reduces the memory footprint of large language models, but preserving accuracy with efficient training remains challenging. Scalar quantization encodes weights independently, while explicit-codebook vector quantization introduces costly search during training. We propose E8-QAT, a 2-bit quantization-aware training framework that uses the $E_8$ lattice as an implicit eight-dimensional codebook. It jointly quantizes weight groups through arithmetic projection without storing or searching an explicit codebook during the training forward pass. A 2-bit core encoding and sparse extension preserve the selected lattice points exactly, adding less than 0.1\% to core index storage. On Qwen3.6-35B-A3B, E8-QAT matches the performance of the full-precision model in terms of the average score across the four tasks. E8-QAT improves over the state-of-the-art baseline by 2.03\% on IFEval and by 1.51\% on LiveCodeBench, under the same data, schedule, and bit width. The 2-bit representation reduces peak memory by 80\% and supports a $4\times$ larger batch. At each representation's largest feasible batch, E8-QAT reaches $1.97\times$ the decoding throughput of BF16.

When Does Experience Memory Help Closed-Loop LLM Driving Agents? A Controlled Evaluation with Resampled Warmup Pools, Token-Matched Fillers, and Audited Baselines

Yingxin Jia, Xuanyi Li — Wed, 21 Oct 2026 00:00:00 +0000

94% · Evaluation 94%

When does experience memory help a closed-loop LLM driving agent? Published gains cannot say, because the two factors we find to matter most, which warmup experiences the pool contains and what any injected tokens do to the prompt, are controlled in none of them. We answer with a controlled evaluation on one audited HighwayEnv testbed (1Hz, gpt-4o-mini for the decisive contrasts). We treat the warmup pool as a resampled random factor (ten disjoint pools) and add token-matched irrelevant-filler controls, oracle-certified diagnostics, caution-pricing metrics, and quantified audits of the baseline ports. The probe suite spans flat-text retrieval (a DiLu port), a single-frame grid port (RESPOND-inspired), our episodic temporal-graph memory, a retrieval-free consolidated rulebook, and a flat K-step observation-history baseline. Four findings follow, each a property of this testbed. (1) Headline memory gains belong to the warmup draw. The reported pool is the best of ten for both retrieval memories, which beat No-Memory in only one pool of ten (sign test p=0.011), and even the rulebook's apparent pool-immunity collapsed on five fresh draws. (2) No retrieval design tested exceeds token-matched filler. The collision-only merge ceiling is reached by a single 15-token caution instruction, and metrics that price caution erase the rulebook's lead. (3) Representational sophistication is nowhere robustly operative. (4) Baseline port quality silently inverts conclusions. On this testbed, a single-pool comparison without filler controls would have reported memory gains that these controls reverse. We release the testbed, audited ports, seeds, and caches.

On-Policy Transition Distillation

Runyang You, Yongqi Li, Tiezheng Yu, Wenjie Li — Wed, 21 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 47%

On-policy distillation (OPD) consolidates expert capabilities into a single student by matching teacher and student final-layer distributions on student-generated rollouts. We investigate whether depthwise prediction transitions, the changes in predictions between layers, can provide additional task-relevant supervision. Our diagnostic study finds that stronger teacher--student agreement in intermediate transitions is associated with higher answer correctness within questions, with correct responses exhibiting lower mean transition KL across all measured intermediate layer pairs. These observations motivate On-Policy Transition Distillation (OPTD), which augments final-layer distribution matching by aligning teacher and student layer-to-layer log-probability increments on student-generated rollouts. Evaluations across three backbones and eight benchmarks spanning mathematics, coding, and instruction following show that OPTD achieves the best or tied-best overall performance in both dual- and three-domain distillation. In the three-domain setting, OPTD recovers $\approx 90\%$ of the overall teacher--student capability gap, compared with $\approx 75\%$ for OPD.

DataPraxis: Scaling Analytics Tasks via Verified Workspace Evolution

Nan Jia, Wenxuan Shi, Jingqing Ruan, Zhizhao Zeng, Tianyi Zhai, Zhongda Su — Wed, 21 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 60%, Hallucination 51%

General-purpose agents are increasingly expected to perform professional analytics work, which requires sustained reasoning, iterative investigation, and reliable evidence. Training these capabilities requires diverse and challenging tasks. However, generating tasks directly from raw data often overlooks deeper questions that emerge through analysis and provides limited evidence for verification. To address these limitations, we introduce DataPraxis, an analytics task synthesis framework based on verified workspace evolution. The framework progressively enriches workspaces with validated findings and artifacts, then inherits and combines these workspaces to uncover deeper questions. It constructs tasks from accumulated findings and their dependencies, verifying analytical results both during evolution and after task assembly. Using this framework, we construct DataPraxis-5K, a dataset of 5,000 verified trajectories spanning diverse domains and output formats, to train DataPraxis-27B and DataPraxis-35B-A3B. DataPraxis-27B achieves 55.11 on DABStep and 51.9 on JobBench, matching or surpassing frontier models such as GLM-5.2. DataPraxis-5K is available at huggingface.co/datasets/stargazerhust/DataPraxis-5k

Teaching AI Agent to Play Poker Strategically Through Preference Optimization

Wed, 21 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 63%

A poker agent can make a legal move and still make a strategically poor one. We ask whether compact AI agents can exhibit strategic behavior under hidden information and whether preference optimization can improve beyond static and zero-shot policies. We first evaluate a static baseline and three frozen language models: SmallThinker, Qwen3, and Gemma-3 whose performance varies substantially between exact-game and six-player evaluation. We therefore propose Strategic Preference Optimization with KL-anchored Equilibrium (SPOKE), combining teacher-relative value penalties, policy anchoring, risk constraints, and structured rationale supervision, and compare it with SFT, DPO, GRPO, and GSPO. Across three runs of a SmallThinker-4B-A0.6B policy, SPOKE achieves the lowest mean exploitability in Kuhn (0.213) and Leduc (1.081) and the highest six-max top-1 agreement (0.775), while maintaining teacher KL comparable to SFT (0.522 vs 0.523). These results provide preliminary evidence that strategic preference optimization can improve strategic quality while preserving teacher-consistent behavior.

Sparse RL Relaxation for LLM Reasoning

Jiabin Fan, Matthias Lindemann, Zita Marinho, Andr\'e FT Martins, Lili Mou — Wed, 21 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 39%

Recent work has explored continuous representations in place of discrete tokens during Chain-of-Thought (CoT) reasoning, often referred to as soft CoT. This is motivated by the idea that a mixture of token embeddings can represent a superposition of multiple reasoning possibilities within a single step. However, it remains challenging to train large language models to realize this potential. Existing work has explored reinforcement learning (RL)-style training for soft CoT. However, these approaches treat soft tokens as actions in a Markov decision process rather than as differentiable intermediate computation steps. Thus, they do not propagate gradients across the soft tokens and miss out on the signal of how the choice of one soft token affects the next. In this paper, we propose a sparse relaxation method for RL (called SpaRLax), which leverages the differentiability of soft-CoT reasoning over time steps and makes training more stable and effective. We adopt an entmax sparsity method that adaptively determines the degree of softness, which truncates long-tail noise at confident steps while retaining multiple plausible candidates when uncertain. Across nine reasoning benchmarks, our method achieves the highest average performance compared with baseline methods. Further analyses validate the effective use of soft-token superposition in our approach, and also demonstrate the contribution of adaptive sparsity in our approach.

Projected Distillation: On Learning to Solve Hard Problems from Close, Correct Targets

Wed, 21 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 38%

On-policy learning is central to post-training: training on a model's own generations limits distribution shift and helps preserve existing capabilities. On hard problems, however, on-policy RL provides no reward-based learning signal when all attempts fail. On-policy distillation provides next-token supervision from a teacher, but on hard problems, if a rollout goes wrong early, subsequent supervision is conditioned on prefixes that already contain the mistake. Correct behavior can instead come from a gold or teacher solution, but such off-policy targets can differ substantially from the model's generations and, as we show, may erode existing capabilities. We therefore treat on-policyness as a continuum and propose projected distillation, which minimally edits the student's failed attempts into correct targets. The method uses contrastive updates to favor corrections over failed attempts and refreshes these corrections as the student evolves, enabling near-on-policy learning even when all student attempts fail. Theoretically, we prove that the student's learning regret grows with the edit size: it scales linearly with the horizon under minimal correction, as in DAgger, and quadratically under full-answer replacement, as in behavior cloning. Empirically, across two students (Qwen3-8B and OLMo-2-7B) and five domains spanning math, code, science, and tool use, projected distillation gives the highest area under the specialization-retention frontier (best on 8 of 9 cells against GRPO and OPD, and 7 of 9 against SPOT and SODA). Our ablations highlight the importance of online corrections, the contrastive loss, and on-policyness: moving targets farther off-policy steadily worsens the specialization-retention trade-off, with retention falling from 0.916 to 0.744 when minimal corrections are replaced by student-agnostic gold solutions.

Judging the Justification: Decision Certificate Supervision for Generative Reward Models

Guiquan Sun, Bing Bai, Martin Renqiang Min, Dongjin Song — Wed, 21 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, New model 86%

Generative reward models (GenRMs) generate natural language rationales to justify preference judgments, yet a correct verdict does not guarantee a sound justification. Existing approaches often rely on reference explanations to supervise rationales, but obtaining such references can be costly and does not necessarily establish whether generated rationales support the resulting verdict. To address these limitations, we propose DeCert, a framework that evaluates the validity and verdict consistency of generated rationales without requiring reference explanations. Crucially, DeCert translates these assessments into preference supervision that accounts for both verdict correctness and justification quality. In particular, when multiple sampled judgments share the same correct verdict, differences in certificate quality enable DeCert to construct rationale preference pairs that distinguish stronger from weaker justifications. Extensive empirical studies across two model families and multiple in- and out-of-distribution benchmarks demonstrate that DeCert consistently outperforms strong generative judge baselines while improving rationale quality. Moreover, DeCert effectively mitigates positional bias, improving position-consistent accuracy by up to 14.8 points. These results highlight the importance of supervising not only which verdict is correct, but also how that verdict is justified.

PAR: Progress-Aligned Recurrence for Length Generalization in State Tracking

Pan Wang, Yihao Hu, Guangyuan Dong, Yitong Lin, Hang Wang, Zimi Liu — Wed, 21 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 48%

Looped language models can run more iterations at test time, but additional iterations need not help them solve longer problems. For sequential state tracking, final-answer supervision does not specify how far each iteration should advance, and repeated updates can change the scale of the recurrent state. We introduce \textbf{Progress-Aligned Recurrence (PAR)}, which trains the prediction at iteration $k$ to match the state after the first $k$ operations and rescales recurrent states to their first-iteration norms. On swap-$S_5$, a Qwen2.5-0.5B model trained on at most 32 moves achieves 95.5\% mean full-state accuracy at 512 moves, or $16\times$ the training length, compared with 55.7\% reported for RLT and 0.8\% for explicit chain of thought on the same backbone. Controlled $S_5$ ablations show that removing norm anchoring reduces the extrapolation factor from $16\times$ to below $2.2\times$, and replacing progress-aligned targets with final-answer supervision reduces it to at most $8\times$. A learned completion readout selects when to answer without a supplied stopping depth. When intermediate labels are unavailable, prefix self-distillation induces the correspondence from shorter problem prefixes using only final answers. The gains apply across sequential state-tracking tasks, establishing the task structure required for depth extrapolation. These findings support training intermediate predictions and state scale together to make additional recurrent depth useful for longer problems.

How Language Models Organize Task Information

Kaiyang Wan, Yuxia Wang — Wed, 21 Oct 2026 00:00:00 +0000

93% · Evaluation 93%

Dense linear probes assess task readability from language model (LM) representations, but do not directly reveal how many features suffice for prediction. We measure readout sparsity with $k^*$, the smallest feature count found by a fixed search protocol that recovers 95% of dense-probe AUROC above chance. We evaluate 19 LMs from five labs on 14 core tasks. Ten base models give the nine reliably readable tasks similar feature counts and nearly the same order. This agreement extends to the selected features themselves, which on three tasks show cross-model response correspondence beyond class-mean differences and a label-correlation-matched null. Instruction-tuned deployment selectively reduces $k^*$, with both weights and input format contributing. In controlled synthetic experiments across 20 models, successfully learned targets admit single-feature readouts, whereas readable controls require tens of features. Reference-model $k^*$ is associated with label requirements across eight students, following an empirical scaling of $N^*\propto(k^*)^{0.65}$. After continual learning, saved heads can fail while linear readability remains high, and recovery requires fewer labels than initial task learning under the tested procedures. Pruning reduces linear readability, with wider pre-pruning readouts generally associated with larger losses. These findings distinguish task readability from readout sparsity and establish their complementary value for predicting learning requirements and diagnosing LM changes.

SliMOO: Interpretable Multi-Objective Evolutionary Search for LLM Depth Pruning

Guanchen Li, Yixing Xu, Xuanwu Yin, Dong Li, Emad Barsoum — Wed, 21 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 63%

As large language models (LLMs) grow more powerful, their scale makes them costly to serve. Depth pruning offers a practical route to efficiency by directly reducing memory and latency without hardware-specific support. However, we observe that transformer layers exhibit markedly different task-dependent layer contributions, whereas existing depth-pruning methods ignore this heterogeneity and thus tend to optimize for a narrow domain or benchmark rather than preserve broad capability. This suggests that LLM depth pruning should be formulated as a multi-objective optimization problem rather than a single-score ranking problem. We propose SliMOO, an interpretable multi-objective evolutionary search framework for LLM depth pruning. SliMOO first uses single-layer removal probing to construct a task-layer importance map, which provides an interpretable view of task-shared and task-specific layers and serves as a search prior. It then performs a Pareto-aware evolutionary search for pruning masks, where candidates are generated under the guidance of the task-layer prior, scored by their deviation from the dense model on each task, and retained through NSGA-II environmental selection. Across the Llama-3.1 and Qwen3 model families, SliMOO consistently achieves stronger multi-task trade-offs than rule-based, greedy, and scalarized baselines. For example, on Llama-3.1-70B at 25% sparsity, it improves the math-domain average from 24.33% to 34.40% and the code-domain average from 35.03% to 39.14%. Code will be made available at https://github.com/AMD-AGI/SliMOO.

Adaptive Stopping of Streamed Reasoning for Language Model Agents

JunHyeok Oh, Byung-Jun Lee — Wed, 21 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Agents 89%, Evaluation 81%

Reasoning-enabled language model agents can incur substantial computation before acting, although the reasoning needed varies across decisions. Existing stopping signals, such as confidence and intermediate-answer stability, can vary in effectiveness across reasoning models; signals obtained from hidden states or token probabilities also require internal access that may be unavailable or impractical. We introduce StreamStop, which learns model-specific stopping policies offline from prefix-level action agreement. A frozen prefix reader encodes available reasoning text for a lightweight stop head, allowing stopping without access to the reasoning model's internal signals. We construct and publicly release a dataset of agent reasoning trajectories with prefix-level action annotations. These annotations compare each prefix action with the full-reasoning reference action, enabling offline comparison of stopping rules and exact expected-return optimization of a discounted action-agreement objective. Conditioning the head on the discount factor allows the agreement-token trade-off to be adjusted at inference without retraining. In closed-loop evaluations on WebShop and ScienceWorld, StreamStop achieves higher observed success than most validation-selected fixed-budget settings while using fewer reasoning tokens than Full thinking. On WebShop with Qwen3.8-27B, it uses 42% fewer reasoning tokens per episode with similar observed task success.

Adaptive Pullback: Reweighting Supervision at Penalty Jumps in On-Policy Distillation

Guangyu Liu, Siyuan Gan, Shilong Zhou, Linjian Meng, Boyan Wang, Jing Huo — Wed, 21 Oct 2026 00:00:00 +0000

93% · New model 93%

On-policy distillation (OPD) supervises student-generated responses using sampled-token student-teacher log-ratios. A positive log-ratio penalizes a token that the student assigns higher probability to than the teacher, but its current value does not distinguish sustained high penalties from abrupt increases along a response. We examine these local changes by comparing consecutive blocks of tokens within the same response. Our diagnostic identifies penalty jumps, abrupt increases in average penalty that persist on held-out tokens. Across these jumps, penalties rise while teacher confidence decreases. We call this pattern penalty-confidence tension. Boundaries into high-penalty blocks do not exhibit comparable confidence changes, showing that local changes provide information beyond the current penalty value. Motivated by this observation, we propose Adaptive Pullback OPD (AP-OPD). At each detected jump, AP-OPD uses the preceding block as a local reference and uniformly scales all supervision coefficients in the following block so that its weighted average penalty matches the preceding block's original mean. This reference determines the scaling factor without an additional strength constant. AP-OPD improves performance on both mathematical reasoning and code generation.

Self-Credit: Credit Assignment for Long-Horizon Agentic RL via Intra-group Rescaling

Wed, 21 Oct 2026 00:00:00 +0000

93% · New model 93%, Agents 90%, Evaluation 37%

LLM agents are commonly trained with RLVR, but the sparse nature of typical RLVR rewards makes credit assignment challenging, especially for long-horizon tasks. We propose Self-Credit, an end-to-end agentic RL method that creates dense supervision signals from sparse outcome rewards using self-generated credit assignment via intra-group rescaling. Self-Credit builds on group-based RL with three components. First, during its rollouts, the policy model generates summaries with citations to earlier turns containing critical evidence, which serve as self-generated credit. Second, we refine this model-generated credit with intra-group rescaling, which compares cited actions across successful and failed rollouts for the same prompt. This approach uses group-level agreement and disagreement to remove undeserved credit and further emphasize credit assigned to critical actions. Third, advantage reweighting uses the rescaled credit weights to update the advantages of the cited turns. On BrowseComp-Plus, Self-Credit improves over the base models by 37.0 and 36.2 points on Qwen3 8B and 32B respectively, and outperforms the strongest agentic RL baselines by 8.5 and 5.6 points, respectively. Self-Credit also improves over the strongest baselines by 4.9 points on GAIA and 4.7 points on xBench-DeepSearch. Analyses show that self-generated credit provides useful signals for selecting critical actions, and that intra-group rescaling, especially on failed rollouts, is important across model scales.

Asymmetric Looped Transformers: Fixed Global Representations for an Iterative Decoder

Cheng Liang, Tengfei Zhang, Wenjie DU, Jianyu Wu, Changlian Ma, Shuchen Pu — Wed, 21 Oct 2026 00:00:00 +0000

93% · New model 93%

Looped Transformers increase effective depth by repeatedly applying shared layers, offering a way to scale computation without increasing parameter count. However, full-stack looping places context construction and iterative state refinement within the same recurrent computation, without explicitly distinguishing their roles. We introduce the Asymmetric Looped Transformer (AsymLoop), which separates single-pass context encoding from recurrent state refinement. A single-pass causal encoder produces layer-specific Global key-value (KV) caches that remain fixed across iterations, while a weight-tied recurrent decoder generates Local KV from its evolving state. Joint attention over both sources under a single softmax allows each update to integrate persistent contextual information with the current intermediate state. For training, we apply a heavy-tail-guided layerwise learning-rate method to the encoder and decoder. Experiments on FineWeb-Edu across four model scales, from 60M to 1B, show consistently lower validation perplexity and higher average downstream accuracy than Full Loop under matched effective depth and training token budgets. At 1B, AsymLoop improves average zero-shot accuracy across seven commonsense reasoning benchmarks from 49.58% to 50.85%. Further experiments extend these benefits to longer training budgets, an OLMo-style backbone, and sparse mixture-of-experts models. These results support asymmetric context encoding and state refinement as an effective design for recurrent language models.

Beyond the Sequentiality Bottleneck: A Hybrid Diffusion-Autoregressive Framework for Efficient Multimodal Reasoning

Wei Cai, Jian Zhao, Yuchen Yuan, Xuelong Li — Wed, 21 Oct 2026 00:00:00 +0000

93% · New model 93%, Efficiency 83%

Autoregressive models (ARMs) face a "Sequentiality Bottleneck" in reasoning, while diffusion language models (DLMs) offer parallelism but lack semantic coherence. To addresses both limitations, we propose Collaborative Reasoning Asymmetric Learning (CoRAL), a hybrid framework via an Asymmetric Planner-Executor (APE) architecture: a DLM Planner generates K diverse reasoning trajectories in parallel, while a large ARM Executor selects the optimal plan and synthesizes the final answer. To train the Planner effectively, we introduce Reward-Guided Denoising for CoT (RDC), which integrates outcome-based rewards into the diffusion process to steer it toward goal-directed reasoning. Experiments on vision-language benchmarks show that CoRAL significantly outperforms prior diffusion-based methods and achieves competitive performance against strong ARMs, achieving both superior efficiency and quality.

Capable Teacher, Myopic Scorer: Reward-Free Teacher Repair for Agentic On-Policy Distillation

Shu Zhao, Tan Yu — Wed, 21 Oct 2026 00:00:00 +0000

93% · New model 93%, Agents 46%, Evaluation 34%

Agentic on-policy distillation (OPD) trains a student agent on its own trajectories while a teacher scores every token. A natural explanation of where it goes wrong is compounding error: the student's mistakes carry its trajectories beyond the region where the teacher is reliable, so recent methods keep training inside that region, through episode-length curricula, teacher-guided turns, prefix replay, or teacher intervention where the two disagree. We show this explanation is incomplete. From the states where the student departs most from the teacher, the teacher's own takeovers still finish the task, yet the student trained on its token scores keeps running out of turns. We attribute this to the myopia of next-token scoring. The teacher rates each token without seeing where the episode ends, so its scores let through the safe, delaying moves that outcome-based RL would punish with zero reward, a penalty agentic OPD lacks. We call this difference between what the teacher does and what it scores the rollout-scorer gap. We propose Rollout-to-Scorer Distillation (R2S), which repairs the scorer before distillation by mining high-disagreement states from the untrained student, running teacher takeovers from them, keeping the states where enough takeovers pass a reward-free check such as consensus or an LLM judge, and fine-tuning the teacher to give its takeover-informed scores on every assistant token from the divergence turn onward, without the takeover in context. The repaired teacher is frozen as the sole scorer for standard OPD, requiring no task reward during repair or training. Across three environments, two teachers and four students, R2S outperforms three state-of-the-art agentic-OPD methods in all twelve teacher-student-environment cells, by 12.8 success points on average over the best of them in each cell. Project page: https://tree-shu-zhao.github.io/r2s/.

AtomSD: Procedural Atom-Guided Self-Distillation for LLM Agents

Bohan Yu, Pengfei Cao, Kai Tian, Chen Han, Shi-Yang Li, Chenxi Zhou — Wed, 21 Oct 2026 00:00:00 +0000

92% · New model 92%, Agents 34%

Large language models (LLMs) possess broad factual knowledge and strong reasoning capabilities, but solving complex tasks also requires procedural knowledge about how to act. Existing approaches often represent procedural knowledge through trajectories or bundled skills, limiting flexible reuse across interaction histories. This paper introduces Procedural Atoms, fine-grained and independently applicable units that represent Intent, When, Do, and Avoid guidance. Building on this representation, we propose AtomSD, an Atom-guided Self-Distillation framework in which the learner continually extracts Procedural Atoms from its own successful trajectories and distills them into the evolving policy through Atom-conditioned supervision and task-reward optimization. Comprehensive experiments show that AtomSD consistently outperforms strong reward-based and distillation methods, exceeding the state-of-the-art baseline by up to 9.31 points in average performance. Further analyses demonstrate that the proposed Procedural Atoms provide effective guidance, highlighting the value of atom-level procedural knowledge for policy learning.

SPEM: Structure-Aware Procedural Memory for Efficient Skill Use in LLM Agents

Ran Zhou, Zehui Kong, Fanyu Lin, Jiaheng Liu, Zhen Zhang, Yuxiang Ren — Wed, 21 Oct 2026 00:00:00 +0000

92% · Efficiency 92%, Agents 61%, Evaluation 30%

LLM-based agents increasingly rely on reusable skills. However, scaling skill libraries makes efficient and accurate skill use difficult. Full skill loading incurs high token and runtime overhead. In contrast, flat semantic retrieval treats skills as independent text entries, ignoring dependencies, execution order, state transitions, and data flow constraints. We propose SPEM, a lightweight structure-aware procedural memory framework for efficient skill use in LLM-based agents. SPEM maintains Atomic Skill Memory for individual skills and Skill-DAG Memory for reusable compositions derived from successful executions. For a new task, SPEM reuses high-confidence Skill-DAG templates when available. Otherwise, it retrieves atomic skills and invokes Skill-GNN for state- and structure-aware skill selection. Successful executions are written back as new Skill-DAGs, allowing procedural memory to accumulate over time. Experiments on ALFWorld and SkillsBench show that SPEM achieves the best overall effectiveness--efficiency trade-off, reaching 87.1% ALFWorld success and 38.38% SkillsBench reward while reducing token and runtime costs against strong baselines.

Exploration via Activation-Space Steering for Reinforcement Learning with Verifiable Rewards

Yunan Zeng, Jinhao Jing, Xi Cheng, Yuwang Wang — Wed, 21 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 32%

Reinforcement learning with verifiable rewards (RLVR) trains a language model by sampling responses, scoring them with an automatic verifier, and updating the policy from the resulting rewards. The policy therefore receives direct learning signals only from trajectories that appear in its rollouts. Yet repeated sampling from the same policy can concentrate on similar reasoning paths, leaving useful alternatives unexplored under a fixed rollout budget. We introduce Activation-Space Exploration (ASE), an online method for diversifying training rollouts. By clustering activation differences between correct and incorrect responses to the same problem, ASE builds a structured bank of latent exploration vectors. Different vectors are then applied to subsequent rollouts to explore alternative reasoning paths. On Qwen3-1.7B-Base and Qwen3-8B-Base trained on DAPO-Math-17k, ASE improves average Mean@8 across six mathematical reasoning benchmarks over GRPO by 4.87 and 1.87 percentage points, respectively, under the same rollout budget.

PORA: Learning When More Reasoning Pays Off for LLM Agents

Bo Yan, Qi Zhang, Zhen Tan, Tianlong Chen, Song Wang — Wed, 21 Oct 2026 00:00:00 +0000

92% · Efficiency 92%, New model 79%, Evaluation 44%

LLM agents perform tasks over multiple steps and revise their actions based on feedback. As the total reasoning budget is often limited and must support future steps, the agent must choose between short and long reasoning before each action. Longer reasoning may improve the final task outcome by preventing later errors, but it also uses budget capacity that may be needed for future decisions. This creates two challenges: estimating the downstream value of long reasoning and deciding when that value is worth using the remaining budget. We propose Paired Outcome-guided Reasoning Allocation (PORA), which learns this decision by comparing short and long reasoning from the same interaction state. During training, \method{} completes the task twice from each selected state, once with a short reasoning allowance and once with a long allowance, and compares their final quality and total processing to estimate when longer reasoning is valuable. During inference, a controller weighs this estimate against learned prices on remaining-budget pressure and chooses either short or long reasoning without running both alternatives. We evaluate PORA across application, memory, code, retail, and mathematics tasks, achieving strong results across all settings. On BFCL memory, it exceeds the strongest existing method by 17.4--32.9 percentage points. Within the same agent, it matches or exceeds always using long reasoning on code and mathematics with 50.6% fewer actor tokens and 31.2% fewer model requests, respectively.

Preference-Aware Pareto Steering for Multi-Task Speech LLMs

Wed, 21 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 48%

Speech large language models (LLMs) perform multiple tasks, including automatic speech recognition (ASR), timestamp prediction (TS), and speech translation (ST). Joint training of speech LLMs creates task trade-offs, yet standard training with fixed loss weights does not generally control the resulting trade-off according to deployment preferences. To address this, we propose Endpoint-Anchored Preference-Aware Sampling Uniformly along the Pareto Front (EA-SURF), an a priori method that adapts scalarization weights so that the resulting task trade-off reflects the desired preference. EA-SURF retains the standard weighted-loss training pipeline, tracking the current model's performance relative to each task-specific reference and adapting the scalarization weights as training proceeds. We theoretically show that EA-SURF can match a desired preference, with the update becoming stationary at the target under checkable regularity conditions. Across six three-objective speech LLM requests, EA-SURF empirically converges toward the requested targets and increases endpoint-inclusive normalized hypervolume by 17.8% over fixed loss weighting.

Ask to Distill: Learning to Synthesize Data for On-Policy Distillation

Jiashuo Sun, Hao Zhang, Xueqiang Xu, Heng Wang, Yixuan Xie, Pengcheng Jiang — Wed, 21 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 50%

On-policy distillation (OPD) can succeed with surprisingly few training questions, motivating a view of questions as carriers of distributional knowledge transfer. How can we assess whether a newly generated question is valuable for distillation, and learn to generate better questions---without access to its correct answer? We introduce Ask-to-Distill (\method), a framework that treats synthetic questions as distribution elicitors and optimizes the teacher supervision they expose along student-generated trajectories. A student-initialized synthesizer is optimized using a distributional surrogate for question value, computed from a frozen student--teacher pair without reference answers, correctness rewards, or per-candidate student updates. The reward combines teacher--student disagreement with teacher probability mass on the student's candidate support, together with length and diversity controls. The generated questions are filtered and deduplicated into a prompt-only bank for standard OPD. Across mathematical reasoning, code generation, and agentic tasks, our method improves average performance and teacher-gap recovery over unoptimized student-generated prompts and KL-filtered real questions, while matching or exceeding real-bank OPD in the evaluated settings. Signal analyses further show that optimized questions can elicit denser supervision and more total signal from shorter student responses, supporting elicited distributional supervision as an objective for synthetic distillation data. Our code, data and models are publicly available at https://github.com/GasolSun36/Ask_to_Distill.

Self-Evolving Agents through Privileged Retrospective Distillation

Yuhang Zhou, Yifan Wu, Bo Peng, Mingyi Wang, Jiayi Liu, Xiangjun Fan — Wed, 21 Oct 2026 00:00:00 +0000

92% · New model 92%, Agents 71%

Self-evolving agents aim to improve from their own experience rather than remain fixed after deployment, but naturally collected trajectories often lack ground-truth solutions or verifier-derived rewards. We introduce Retrospective Self-Distillation (RSD), a multi-round framework that turns such experience into dense supervision using privileged retrospective context. For each on-policy trajectory, the same underlying model acts as both student and teacher: the teacher re-scores the student's realized actions while conditioning on cross-task precedents and lessons, together with turn-level reflection and future hindsight from the completed trajectory. The resulting teacher distribution is distilled back into the policy through on-policy self-distillation. Across rounds, the experience bank grows and the retrospective context is refreshed, allowing the model, its data, and its supervision to co-evolve. We further introduce frontier-guided context-policy evolution, where a stronger frontier model proposes constrained edits to how privileged context is constructed. Across ALFWorld, ScienceWorld, and SearchQA with Qwen3-8B and Qwen3-14B, RSD improves performance by up to 59.7% relative to the base model, while frontier-guided context-policy evolution adds up to a further 44.7% relative improvement over a fixed context policy. These results show that naturally accumulated experience can be converted into persistent model capability without ground-truth trajectories or verifier-derived rewards for the core policy update.

How Fine-Tuning Objectives Shape Layer-Wise Information in LLM Hallucination Detection

Kaiyang Wan, Forrest Sheng Bao, Amin Ahmad, Yuxia Wang — Wed, 21 Oct 2026 00:00:00 +0000

92% · Efficiency 92%, Hallucination 85%, Evaluation 30%

Under matched conditions, classifier-head fine-tuning (\textsc{cls}) paradigm consistently outperforms the LLM-as-Judge sequence-to-sequence (\textsc{s2s}) on binary hallucination detection. We trace this gap to the representations induced by the two objectives. Static analysis shows that \textsc{cls} learns a task-specific discriminative direction, pushing hidden states into a clear two-cluster geometry with stronger class separation and linearly decodable label information. By contrast, \textsc{s2s} relies on the pretrained language-modeling head to make decisions. This minimizes the pressure to reorganize representations, leaving the hidden states much closer to the pretrained backbone. Our analysis also reveals that the \textsc{cls} signal saturates before the final layer, after which upper layers form a plateau of statistically interchangeable and largely redundant representations. However, endpoint redundancy alone does not show whether these layers are dispensable during training. Further training-dynamics analysis reveals 1) a lower pretrained substrate, 2) a narrow upper-middle \textbf{InfoWindow} where task information emerges, and 3) a later-synchronizing upper plateau that mirrors this signal. This mechanism motivates us to remove layers above the InfoWindow. Truncating 31\% of the network leaves accuracy unchanged within seed noise on Qwen3-4B and 8B and cuts latency by $1.4\times$. The same depth fraction, fixed before any analysis of the target models, transfers to Llama-3.2-3B and Mistral-7B-v0.3, and the layer structure behind it reappears there and on QNLI and SST-2. We hope this helps guide efficient task-specific LLM design.

TrajEvolve: Trajectory Evolution for Reinforcement Learning with Hindsight Credit Assignment

Yi Wen, Hao Chen, Wanyu Wang, Maolin Wang, Pengyue Jia, Derong Xu — Wed, 21 Oct 2026 00:00:00 +0000

92% · New model 92%

The Reinforcement Learning (RL) paradigm has achieved remarkable success in enhancing the reasoning capabilities of Large Language Models (LLMs). However, when applied to long-horizon tasks, it faces severe challenges: i) Due to the high task complexity, the model struggles to sample enough successful trajectories, resulting in slow convergence. ii) Relying solely on sparse final rewards causes all steps to be updated indiscriminately, encouraging redundant step generation. When both the quantity and quality of successful trajectories become difficult to guarantee, the performance is severely constrained. This dilemma stems from the limitations: existing RL paradigms rely entirely on the reasoning capability of base models, lacking the ability to actively optimize trajectories. To address the above issues, we propose a novel trajectory evolution paradigm with hindsight credit assignment, termed TrajEvolve. TrajEvolve not only relies on the trajectory inferred by the model but also generates ``evolutionary trajectories'' by refining low-quality steps in historical trajectories, thereby exposing the policy to high-quality compositions that lie within the base model's support but are rarely sampled within practical training budgets. Furthermore, TrajEvolve reformulates step-level continuous credit assignment as a binary classification task, whether each step is necessary for task completion. This simplification not only reduces the learning difficulty but also enables step labels to be obtained through counterfactual verification, achieving precise identification of redundant steps. Extensive experiments on two datasets demonstrate the superiority of the proposed method.

Same Judgment, Different Yes: Measuring Multilingual Judgment Profiles in Language Models

Haonan Huang, Yuwei Lu, Chengkun Cai — Wed, 21 Oct 2026 00:00:00 +0000

92% · Evaluation 92%

Multilingual evaluations often score a model's opening yes or no, a token that also carries model and language answering conventions. We ask when different ways of eliciting and reading an answer recover the same judgment, and when they measure an answer habit. We collect 1.1 million replies from eight models in six languages on up to 105 moral vignettes, varying the question's form, the answer constraint, and the reading of the reply, token or sentence, in both wordings (the FAR-6 battery). Story-level judgment profiles converge across the tested readouts and question forms, except for the yes/no token after a negative interrogative read under a preset mapping: on the same replies, token and sentence profiles agree at a median r = 0.67 across cells, while sentence profiles agree with free-sentence profiles at 0.94, a known linguistic variability. With each language's story consensus removed, a model's profile is nearest to its own in another language for 39 of 48 model-language cells under proposition labels and 15 under free sentences, against 5 expected by chance. Particle, position, and named-term preferences follow their own patterns. A model can keep its judgment and change its yes: scoring that token requires establishing which proposition it affirms.

TPO: Tri-level Distributionally Robust Learning for OOD Direct Preference Optimization

Chengtao Jian, Kai Yang — Wed, 21 Oct 2026 00:00:00 +0000

92% · New model 92%

Direct Preference Optimization (DPO) minimizes empirical risk over observed preference comparisons, implicitly assuming that the training preference distribution matches the deployment preference distribution. In practice, this i.i.d. assumption often fails, as preference distribution shift exhibits a natural hierarchical structure across three levels: group-level shift, prompt-level shift, and preference-level shift. Such multi-level distributional mismatch poses a critical challenge for safe LLM deployment, yet existing robust DPO methods typically either model uncertainty over the joint preference distribution or focus on individual sources of shift, leaving the hierarchical structure of OOD preference shifts underexplored. We propose TPO, a tri-level distributionally robust framework for OOD preference alignment, which jointly models all three levels of distributional uncertainty through a hierarchically structured rectangular ambiguity set. By instantiating it with Wasserstein distance, we derive a tractable and interpretable objective that combines a closed-form preference-level margin penalty, prompt-level adversarial embedding perturbation, and worst-case group reweighting, each governed by an independent radius. We establish finite-sample convergence guarantees and empirically validate the effectiveness of TPO over existing robust DPO methods.

Forty Questions: Identifying a Language Model from One-Word Answers

Tapan Parikh — Wed, 21 Oct 2026 00:00:00 +0000

92% · Evaluation 92%, Efficiency 50%

A model behind an API can be identified from its one-word answers to everyday questions such as "Name a sound" or "Name a poet". A model's profile records how often it gives each answer to each question. We attribute new answers to the model most likely to have given them. On a panel of 94 models, 40 answers identify the exact model 92.5% of the time, against 1.1% by chance. Profiles built through OpenRouter identified all ten models served by DeepInfra under blinded labels. A system prompt lowers accuracy by 6 to 16 points. We applied the method to 96 endpoints of 13 open models on OpenRouter, and 92 identify as the model they claim. Of the other four, one matches the claimed model with reasoning turned off. One serves Gemma 4 under the name Llama 4 Maverick: it names itself Gemma 4, and it is billed Gemma 4's token counts on every message. Two remain unresolved. Billing records also show context added to first-party models. Grok 4.7 is billed 1,247 input tokens for a one-line question that costs GPT-4o only 12 tokens, and five models state the current date without being billed for it. The questions come from a public census of model defaults, and the method costs cents per endpoint.

Beyond the Point Mass: Acceptance-Derived Weights and Distributional Losses for Speculative Decoding

Tianyu Wu, Zhuohan Wang, Jonathan Geuter, Yu Yao, Weifan Jiang, Chengxi Zhang — Wed, 21 Oct 2026 00:00:00 +0000

92% · Efficiency 92%, New model 89%, Evaluation 44%

In speculative decoding, a lightweight draft model proposes tokens that the target model verifies in parallel, speeding up generation without changing the target distribution. D-PACE (Wu et al., 2026) trains such drafters with a loss that weights each position by its contribution to the expected accepted length. Both its weights and its cross-entropy (CE) loss use only the draft's probability of the recorded target token, treating the target as a point mass. Under rejection sampling, however, a sampled draft token is accepted with probability equal to the overlap of the draft and target distributions, one minus their total variation (TV). We derive D-PAL (Dynamic Position-wise Acceptance Loss) from the exact expected accepted length, using this overlap as the acceptance signal. This yields overlap-based position weights with log acceptance (LA) as the per-position loss, and the objective reduces to D-PACE for a point-mass target. Since the weights are detached, any per-position loss can replace LA in practice. We show that forward KL divergence and R\'enyi divergence of order 1/2 match or exceed LA, and that R\'enyi-1/2 provably bounds the true acceptance rate from above and below. Replacing CE with R\'enyi-1/2 improves acceptance length under both D-PACE's weights and ours. On six benchmarks with a Qwen3-4B target, D-PAL raises mean acceptance length over D-PACE from 4.50 to 4.57 at T=0 and from 4.20 to 4.28 at T=1, and consistently achieves higher throughput at both temperatures.

Toolflation: Seller Agents Distort Tool Markets Through Spontaneous Generative Engine Optimization

Zhenwei Tang, Di Bai, Jintao Liu, Xidong Wu, Qianfeng Wen, Xinwu Cheng — Wed, 21 Oct 2026 00:00:00 +0000

92% · Agents 92%, Evaluation 67%, Hallucination 47%

Agents are starting to sell to other agents. Seller agents write the listings of tools, services and products, and buyer agents read these listings when they choose what to use. We show that seller agents told only to be selected more often practice generative engine optimization on their own, and that this distorts the market for sellers and for buyers. We build tool markets from real data tools on Amazon and Instagram and run every tool on every question, so we know which tools can answer each question. Under all five production models we test, sellers rewrite their descriptions to claim more and more fields, including fields that their tools return for fewer than half of the targets. We call this process toolflation. Nothing in the market limits it, and how far it goes depends on the model, as the number of claims that contradict the sellers' own evidence differs by an order of magnitude between models. Toolflation shifts the buyers' choices between sellers, and the most capable tool loses share in seven of the ten combinations of seller model and platform. Under four models, the choices stay with capable tools and buyer agents keep their task success. Under the fifth, sellers copy one another until the descriptions no longer tell the tools apart. Buyer agents then pick tools by their position in the list, tools that answer almost nothing win a quarter to two fifths of the choices, and task success falls from 89% to about 30% on Amazon, close to random choice, and from 77% to 40% on Instagram. The sellers did only what their vendors asked, and the cost fell on buyers they never dealt with. We therefore argue that honesty toward third parties under competition is part of safety alignment, and that models should be tested as sellers before they are deployed as sellers.

Navalia: Advancing Cowork Agents through Verifiable Task Synthesis and Long-Horizon Post-Training

Hengqi Liu, Ziyuan Liu, Yang Qin, Wanting Zhou, Zichuan Wang, Jiachen Liang — Wed, 21 Oct 2026 00:00:00 +0000

91% · Agents 91%, New model 90%, Evaluation 50%

Large language model (LLM) agents are evolving from conversational assistants into Cowork Agents capable of autonomous planning, tool use, and complex task execution. Financial analysis requires these agents to integrate cross-document evidence and perform multistep numerical reasoning, but expert task authoring is costly, while direct LLM synthesis struggles to produce challenging tasks with reliable answers. We introduce Navalia, a framework for Cowork Agents that combines verifiable task synthesis with long-horizon post-training. Using corporate financial reports as its data source, Navalia semantically selects supporting evidence from complete fact sets and constructs executable computation chains to generate analytical tasks with traceable evidence and actual data dependencies between steps. We further use independent LLM review and programmatic checks to verify evidence consistency and answer correctness. The verified tasks are then used for trajectory collection, cold-start supervised fine-tuning (SFT), and reinforcement learning (RL) to improve long-horizon task execution. Navalia, initialized from Qwen3.6-35B-A3B, achieves consistent gains over the base model across all five benchmarks: OfficeQA Pro, FinSearchComp, Finance Agent Benchmark, APEX-Agents-AA, and Workspace-Bench-Lite. These results demonstrate that Navalia improves evidence integration, tool use, and long-horizon task execution capabilities and generalizes to broader Cowork settings, including general office tasks.

Predicting Linear Relational Decodability Without the Relation

Wed, 21 Oct 2026 00:00:00 +0000

91% · Evaluation 91%, New model 33%

Linear relational embeddings (LREs) approximate a language model's subject-to-object mapping with an affine function. Their faithfulness varies enormously across relations, from near zero for person_father to the ceiling for country_capital_city. We investigate the factors that contribute to this variation and demonstrate that it can be significantly predicted from a readout that is not aware of the underlying relation. To do so, we propose a novel form of the Jacobian lens, the entity lens. Whereas the Jacobian lens is estimated across all tokens and concepts, the entity lens EL_t is trained to predict what the model will output for an entity of type t under the neutral prompt "{} is known for"; e.g., we train ELs for countries and persons. After estimating an entity lens, we measure how highly it ranks the model's answer for a relation without using any relation-specific information in estimation and prediction. Across 28 relations of the LRE dataset on Olmo-3-7B this relation-agnostic score correlates with LRE faithfulness at \rho = 0.7355 (p < .01). This suggests that linear decodability is largely a property of what a subject's representation already carries, rather than of the relation being decoded. Moving the unit of analysis from the relation to the individual fact, we conclude that new relation datasets and analysis methods are needed to address several problems in existing datasets, most prominently semantically weak first tokens (e.g., most predicted universities' first token is "University") and reference competition (e.g., the LLM confuses obscure actor names with competing non-actor entities that have the same last name). Our full pipeline is available at https://github.com/jonasrohw/entity-lens

AudiPsy: Evidence-Grounded Multimodal Reasoning over Spoken Counseling Conversations for Depression Assessment

Chuanbo Zhu, Run Luo, Weibo Zhao, Yue Li, Kun Qian, Qiang Qu — Wed, 21 Oct 2026 00:00:00 +0000

91% · New model 91%, Evaluation 70%

Depression assessment from counseling speech requires integrating textual and acoustic psychological evidence across turns, yet existing end-to-end classification approaches lack explicit multimodal evidence interpretation and context-aware reasoning. We propose AudiPsy, an evidence-grounded omni-LLM for depression assessment over spoken counseling conversations. AudiPsy employs an Observation-Assessment-Summary (OAS) paradigm to integrate dialogue transcripts, textual and acoustic emotions, and textual-acoustic emotion incongruence into structured reasoning. OAS incrementally organizes multimodal observations, performs contextualized assessment, and maintains an evolving dialogue summary for subsequent turns. To support this framework, we construct AudiPsy-Counsel, a synthetic spoken counseling conversation corpus with multimodal context augmentation, and develop a unified three-stage curriculum for depression-oriented reasoning. Experiments on DAIC-WOZ demonstrate superior depression assessment performance over specialized depression detection methods and general-purpose speech/omni-LLMs, along with improved reasoning quality, while supplementary evaluation on EATD-Corpus supports AudiPsy's applicability to naturalistic conversations.

COMPASS: Routing via Capability-Oriented Model Profiling with Adaptive Search, Selection, and Tool Use

Amin Rezaei, Pouya Parsa — Wed, 21 Oct 2026 00:00:00 +0000

91% · Efficiency 91%, Agents 81%, Evaluation 48%

LLM routing aims to select a capable model at low cost, but specialized routers can require updates as models and tasks change. We propose COMPASS, an agentic router in which an LLM follows a routing recipe, gathers performance and cost evidence through tools, and optionally inspects candidate answers before selecting a model. Our evaluations on LLMRouterBench and DeepSWE show that agentic routing can improve accuracy and lower model-call cost relative to specialized and trained routing baselines, without routing-specific training. On LLMRouterBench, COMPASS reaches 69.85% accuracy at \$139.57, compared with RLCascadeRouter's 69.14% at \$180.22. On DeepSWE under fixed five-fold cross-validation, it reaches 78.1% at \$215, compared with 74.1% at \$500 for the best single model. COMPASS also supports routing across benchmarks and adapts to an expanded model pool by reading new evaluation records, without changing the router's weights or prompt.

Don't Skip the Baseline: A Controlled Comparison of Dynamic Expert Skipping in MoEs

Wed, 21 Oct 2026 00:00:00 +0000

91% · Evaluation 91%, Efficiency 45%

A growing body of work proposes dynamic expert-skipping methods for Mixture-of-Experts (MoE) models that adjust the number of experts activated per token during inference, aiming to preserve model quality while further reducing the number of active parameters. These methods have added calibrated importance scores, inter-expert similarity metrics, or learned per-token allocators in order to identify experts to skip. However, different calibration regimes, different reductions in the effective number of experts, and evaluation suites make comparisons difficult. We classify four representative prior methods into a unified taxonomy and benchmark them against a natural, simple baseline: statically reducing the number of experts $k$ (i.e., same reduced $k$ across all layers and tokens). We provide a fair comparison of these methods at matched expert-compute budget, on two modern MoEs (Gemma4-26B-A4B and Qwen3-30B-A3B) in a text-only setting and a contemporary vision-language model (Qwen3-VL) in a multimodal setting. Surprisingly, we find that the simple baseline is hard to outperform consistently across models, modalities, benchmarks, and calibration settings. The methods we benchmark often underperform the baseline and achieved gains, when present, are minor. We further find that the effective $k$ achieved (or optimized for) in calibration can diverge from the effective $k$ in testing, resulting in over- or under-estimating the compute at inference time. Based on these findings, we recommend a calibration-free per-modality static $k$ reduction and validate this on a held-out model and modality. We further argue that a static control at matched measured compute is the baseline any dynamic skipping method must beat.

Beyond 4 bits: Data-Centric Post-training Quantization and Recovery

Rhea Sanjay Sukthanker, Riccardo Grazzi, James Hensman, Pashmina Cameron — Wed, 21 Oct 2026 00:00:00 +0000

91% · Efficiency 91%, New model 50%

Post-training quantization (PTQ) compresses the weights of a trained model with limited compute and data. Down to 4 bits it is near-lossless, but at 2 bits scalar PTQ configurations fall to chance-level accuracy. We study three choices usually treated as implementation details that matter little at 4 bits but decide 2-bit quality: the quantization representation, the calibration data, and the recovery fine-tuning data, across six dense and mixture-of experts models (ranging from $3.8$ to $32$B parameters). Representation is the largest factor and vector codebooks restore usable models where scalar PTQ pipelines collapse. Most of the remaining loss in accuracy is on \emph{generative} tasks, and the amount of performance drop is model-dependent. Furthermore, at a short recovery budget, calibration and recovery data are coupled. The best recovery data depends on the calibration data, so the two must be chosen jointly. The resulting recipe involves a vector codebook representation, a small task-aligned calibration mixture, and an inexpensive codebook-only recovery stage. Including codebooks, it uses about $2.3$ effective bits per weight and retains $89$ to $97\%$ of full-precision question-answering (QA) accuracy where scalar PTQ is at chance, and, including recovery, uses $1.4$ to $30\times$ less compute than the strongest quantization-aware-training (QAT) baseline.

Pareto-Optimal RTL Code Generation via Multi-Objective Reinforcement Learning with Large Language Models

Wed, 21 Oct 2026 00:00:00 +0000

90% · New model 90%, Evaluation 59%

Recent studies have proposed reinforcement learning with verifiable rewards (RLVR) methods that enable LLMs to generate high-quality RTL circuits from natural-language specifications. Existing approaches optimize circuit PPA (Power, Performance, Area) by aggregating multiple objectives into a single scalar reward, making it difficult to simultaneously generate diverse Pareto-optimal solutions across different trade-offs such as low power, low latency, and small area. However, in practical design, the relative importance of each objective varies depending on product requirements and design constraints; thus, obtaining a diverse and high-quality candidate set is beneficial in that designers can choose a preferred trade-off a posteriori. In this work, we are the first to propose a learning method for LLM-based circuit generation that optimizes, not the quality of a single circuit, but the quality and diversity of the entire circuit set obtained through multiple generations in the PPA objective space. The proposed method consists of (1) a generation-based training pipeline that reformulates coverage improvement of the entire Pareto solution set into a sequential circuit generation problem, and (2) a reward function based on Hypervolume Contribution (HVC) that measures how much a generated circuit contributes to the Pareto front. Evaluation on the RTLLM v2 benchmark shows that the proposed method achieves +13.3 percentage points in relative Hypervolume over a baseline without PPA optimization and +14.4 percentage points over a single-objective reinforcement learning method, while also demonstrating superiority in circuit diversity.

Expert-Guided Rubrics as Dense Reward Signals for Reinforcement Learning in Non-Verifiable Domains

Wed, 21 Oct 2026 00:00:00 +0000

90% · New model 90%, Agents 68%, Evaluation 63%

Large language models (LLMs) are increasingly deployed as autonomous agents in non-verifiable domains such as financial services, where they must interpret user intent, rewrite queries, and retrieve information under domain-specific constraints. Although reinforcement learning (RL) can train such agents from the consequences of their actions, recent successes largely rely on verifiable tasks with automatically computed binary rewards. In non-verifiable retrieval tasks, assessing query quality requires scarce subject-matter expert (SME) judgment that is inherently graded rather than binary. With Group Relative Policy Optimization (GRPO), sparse rewards yield no useful learning signal when all sampled outputs within a group receive the same outcome, causing optimization to stall. We introduce a method that converts a small, fixed investment of SME judgment into a dense reward and combines it with retrieval-based metrics to form a hybrid, GRPO-compatible training signal. Experiments show improvements of up to 6\% in Recall@3 and 6.7\% in MRR@3 compared with sparse-reward baselines.

A general-purpose System One decision model for clinical records

Wed, 21 Oct 2026 00:00:00 +0000

90% · Efficiency 90%, New model 83%, Evaluation 49%

Clinical care, health-system operations and biomedical discovery depend on extracting reliable answers from patient records, yet applying diverse clinical questions across entire populations remains computationally expensive. Here we present Clinical System One (CS1), a 4-billion-parameter decision model for fast, typed judgments from clinical records, along with two agentic harnesses in which CS1 works together with a generalist language model. CS1-4B is trained on 12.4 million questions and answers from 3.6 million patients across a multi-hospital health system, combining structured-data keys, observed outcomes and language-model supervision across clinical notes, diagnostic reports and patient timelines. It answers multiple typed questions in a single forward pass without task-specific retraining, and each answer carries a probability that can route uncertain cases to human review. On 963,408 held-out patients, CS1-4B outperformed open decision models up to 27 billion parameters on 13 of 14 tasks scored independently of its training teacher. Across 21 external data sets, its aggregate performance approached that of the strongest tested 27-billion-parameter model, with 6.4-fold faster inference in a matched workload. Integrating CS1-4B with a language model agent produced cited answers over complete multi-year patient histories and screened a patient population against the eligibility criteria of 17 registered clinical trials, placing four times as many judged-eligible patients among its top 20 candidates as keyword search while using 690-fold fewer language-model tokens than a language model reading every record. These findings support a division of labour in clinical AI, with small decision models performing repeated record reading and generative models formulating questions and synthesizing evidence.

Hybrid Momentum Methods for Optimizing Deep Learning Models

Haifeng Xu, Yijin Ren, Qi Deng, Chenyu Zhou, Dongdong Ge, Yinyu Ye — Wed, 21 Oct 2026 00:00:00 +0000

90% · New model 90%

Momentum is central to stochastic optimization in deep learning, including the training of large language models. Classical momentum methods typically rely on a fast exponential moving average (EMA) scheme, which discounts past gradients on a fixed timescale and may underweight older gradients that remain informative. This creates a trade-off between retaining long-term history and adapting to recent changes. We introduce Hybrid Momentum (HyM), which combines a linearly weighted average of past gradients with a geometrically discounted fast momentum. The former assigns weights proportional to the gradients' iteration indices and retains a growing span of history, while the latter emphasizes recent gradients. This hybrid construction is modular and can be combined with adaptive preconditioning and other update transformations. In particular, we propose HyMUON, which applies Newton--Schulz orthogonalization to the hybrid direction. Illustrative examples and controlled experiments demonstrate the benefits of combining gradient information across these two timescales. Across image classification, machine translation, and language modeling, HyM and HyMUON match or improve upon strong baselines. HyMUON improves over Muon on IWSLT14, and in both the 100k- and 600k-step GPT-2 124M OpenWebText regimes. For LLaMA-1.3B pretraining on C4, HyMUON achieves a final validation loss of 2.777, compared with 2.903 for Muon.

Adaptive Replay Refinement: Allocating Test-Time Compute Between Revision and Resampling

Wed, 21 Oct 2026 00:00:00 +0000

90% · Efficiency 90%, Evaluation 56%

Test-time scaling allocates additional inference compute to refining existing solutions or generating new ones. This raises a fundamental question: Should I keep improving this solution, or stop working on it and try a completely new solution? Existing methods make this decision using fixed refinement schedules, self-verification judgments, or learned controllers, which can waste compute or rely on signals that may be incorrect or require additional training. We introduce Adaptive Replay Refinement (ARR), a training-free method that detects when a stationary revision reaches a fixed point and reallocates the released compute to a new independent sample; replay decoding treats the current solution as a draft and verifies unchanged tokens in parallel, enabling efficient fixed-point detection. Across seven models from five families and seven benchmarks, ARR matches majority voting on three of five mathematical reasoning benchmarks at equal compute, while reducing critical-path serial passes by up to 2.4 times. On MATH500, ARR achieves 84.6% accuracy with only 41% of the serial passes required by majority voting. These results demonstrate that detecting refinement exhaustion enables adaptive reallocation from revision to independent sampling, improving inference efficiency while highlighting the limits of self-revision. Code is available at https://github.com/AhsanBilal7/adaptive_refine.

Recommending Bacterial Cultivation Conditions from 16S rRNA

Juntang Wang, Hao Wu, Yihan Wang, Runkun Guo, Shixin Xu, Dongmian Zou — Wed, 21 Oct 2026 00:00:00 +0000

90% · Evaluation 90%, New model 75%

Recommending bacterial cultivation conditions from sequence data is a challenging structured problem characterized by missing standardized benchmarks and limited use of sequence structure. To address this, we introduce DSNI, a benchmark of 65,023 bacterial strains with phylogeny-aware splits, pairing 16S rRNA sequences with four cultivation conditions (media, temperature, pH, oxygen). Exploiting the mosaic structure of 16S rRNA, where hypervariable regions support discrimination and conserved regions support primer design, we propose a Variability-Gated Encoder that learns position-dependent importance weights conditioned on biological features (region type, position, and local complexity). The gating module adds minimal overhead (1,280 parameters, <1%) while providing an explicit inductive bias valuable in low-data regimes. Experiments demonstrate that the learnable gate improves flat Transformers by 4.7% in joint accuracy and complements RNABERT, yielding 44.2% exact-match joint accuracy. The model also recovers 90.2% of ground-truth cultivation protocols within its top-25 recommendations, narrowing the exhaustive search space by approximately 60 times.

S-Agent 2.0: Spatial Reasoning with a General Coding Agent

Rui Gong, Ruoyu Wang, Yalun Dai, Hao Li, Sidi Yang, Qihao Zhao — Wed, 21 Oct 2026 00:00:00 +0000

90% · Agents 90%

Spatial reasoning asks a model to work out where objects are in 3D space, how they relate to one another, and how they move. Agentic VLMs have recently brought large improvements on such tasks by letting the model actively gather evidence with 3D perception tools instead of answering in a single pass. Even so, on many questions the tools return the right evidence and the model still does not arrive at the correct answer: the gap between model and tools, where raw readings become usable conclusions, is left unfilled. We present S-Agent 2.0, a spatial reasoning agent that equips an off-the-shelf general coding agent with a command-line perception toolkit, and we design the interface around one principle: help the VLM close the evidence loop: acquire scene evidence, remember it in a persistent scene memory, and interpret each reading in plain spatial language. This design keeps the full flexibility of a coding agent, and every reading it acts on stays grounded in the tools' geometry. On MMSI-Bench and ReVSI, the toolkit improves a strong open-source model, Qwen3.8-27B, by +13.9 and +20.5 points over the same agent without tools, and a strong closed-source model, Gemini-3.8-Flash, by +11.2 and +9.6 points. With the open backbone S-Agent 2.0 reaches 66.5 on MMSI-Bench and 82.3 on ReVSI, ahead of the specialized system SpatialClaw on the same backbone (62.6 and 74.5), and with Gemini-3.8-Flash it reaches 74.1 and 90.5 against its 67.0 and 79.7. Both tool gains shrink as the backbone grows more capable, but SpatialClaw's shrinks far more on ReVSI, from +13.2 to +2.7, while ours stays at +9.6. S-Agent 2.0 also works across different general coding agent harnesses, such as Codex and Claude Code.

Sage-A1.5: From Long-Horizon Execution to Scientific Exploration with Capability Composition

Wenjie Lou, Tianshuo Peng, Shiyang Feng, Shengji Tang, Runmin Ma, Ruizhe Chen — Wed, 21 Oct 2026 00:00:00 +0000

90% · New model 90%, Agents 70%

In this report, we introduce Sage-A1.5, a 35B Mixture-of-Experts agentic model that builds on A1 to advance long-horizon problem solving through adaptive capability composition. Compared with A1, Sage-A1.5 learns not only to solve individual tasks, but also to organize work and revise its approach during execution. Sage-A1.5 highlights three improvements: agentic orchestration, verification, and open scientific discovery. First, agentic orchestration extends single-agent execution through task delegation and parallel collaboration among subagents with isolated contexts, enabling complex tasks to be decomposed and addressed concurrently and supporting efficient execution while reducing interference across concurrent subagents. Second, agentic verification strengthens long-horizon reliability through progress checking, evidence validation, and reflective strategy revision, allowing the agent to recognize unproductive search paths and pursue alternative approaches. Third, open scientific discovery extends agentic capabilities beyond answering predefined questions to conducting exploratory research, connecting scientific inquiry with interactive analysis and inspectable research artifacts. We showcase these capabilities through complicated agentic tasks and multidisciplinary scientific applications. Besides, we hope Sage-A1.5 provides the community with an accessible foundation for open scientific discovery, offering a practical path from long-horizon task execution to collaborative, verifiable, and inspectable scientific exploration.

★ GardenBench: A lightweight, daily evaluation of LLM capabilities

Tue, 20 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

We introduce GardenBench, a benchmark with 11 test suites that evaluate the capabilities of Large Language Models (LLMs) across a range of tasks. By systematically prompting LLMs with various lightweight queries, we document day-to-day changes in model performance, including run costs and times, that other benchmarks miss. This allows us to compare the accuracy and features of responses across model providers and versions. We account for daily variations in model accuracy, capturing both model progress and degradation, for instance, when a model is changed post-deployment or when a new model is released. We base our evaluation on general-use tasks and make our results publicly accessible via a live dashboard, diversifying existing approaches to LLM testing. Our main contribution is an innovative, lightweight, and inexpensive benchmark, designed to be consumer-facing and for everyday tasks.

★ CME-SpectrumBench: Can LLMs Analyze Condensed Matter Spectral Data?

Tue, 20 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 44%

Large Language Models (LLMs) are presently well assessed by expert-level mathematics and physics benchmarks which focus on theoretical, symbolic reasoning. However, we lack experimental benchmarks that evaluate data analysis in frontier research problems, a capability that future domain-specific foundation models must have. To this end, we present CME-SpectrumBench, an expert-designed benchmark consisting of 190 questions that measures how reliably frontier LLMs analyze research-level spectral scientific data in condensed matter experiments (CME). Using simulated CME spectral data, we prompt models to identify structures from 2D intensity arrays in the face of noise and limited instrumental resolution, a test of dense context reasoning. Models are scored by how accurately their numerical answers approach the ground truth. Under direct query, scores generally pale in comparison to handwritten code due to clear failure modes: models can find simple features but struggle to trace patterns or compute derived quantities, and performance typically degrades with increased data resolution and noise/convolution levels. LLM performance becomes comparable to handwritten code by allowing tool use (access to a Python interpreter) with/without ReAct prompting, in effect mapping questions onto agentic code generation problems. Our findings show that frontier models can serve as research assistants by interfacing with data through code, but still face difficulties processing information directly from large datasets - a key barrier future foundation models must overcome. The code and data are publicly available at https://huggingface.co/datasets/jingene/CMESpectrumBench.

★ ConCI: A Controlled Counterfactual Benchmark for Contextual Integrity in LLMs

Juhyeon Park, Sumin Yu, Taesup Moon — Tue, 20 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Contextual integrity (CI) holds that the appropriateness of sharing personal information depends on the context of the information flow, not on the information alone. However, existing CI benchmarks often use independent scenarios that vary multiple CI parameters simultaneously and are dominated by inappropriate flows, making it unclear whether correct predictions reflect sensitivity to context or a general refusal tendency. We present ConCI, a controlled-counterfactual benchmark in which an anchor flow is paired with two counterfactuals that each modify a single CI parameter. The resulting triplet is scored jointly, with credit given only when all three flows are judged correctly. Built from six existing benchmarks, ConCI contains 1,592 flows in 556 triplets, is approximately label-balanced, and is validated by seven human annotators per flow. Across seven LLMs, many correct anchor predictions fail on at least one counterfactual, and even the strongest model errs on about a fifth of triplets. Furthermore, models perform near ceiling on inappropriate flows but weaker on appropriate ones, often continuing to judge a flow inappropriate even when a contextual change should make it appropriate. Larger models narrow this bias toward non-disclosure but do not eliminate it, and reasoning yields only modest gains.

★ HABIT-Bench: Evaluating Habit Induction from Longitudinal Weak Evidence in Agent Memory

Junming Zhang, Rubin Wei, Xu Cheng, Hao Sun, Ruoxi Cheng, Xiquan Wang — Tue, 20 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

In long-term interactions with large language model (LLM) agents, users reveal habits through their choices and actions. These habits capture implicit user information that agents must infer across conversations. However, individual observations may provide only partial evidence about a habit or when it applies. Retaining past interactions alone does not ensure that agents can infer habits and apply them appropriately. We introduce HABIT-Bench to evaluate how well agent memory systems infer habits from weak evidence and apply them to later requests. We construct synthetic user histories from task-oriented dialogue data and controlled habit graphs. The histories span hundreds of sessions across food, finance, software, and travel. Four-choice tasks assess agents' ability to infer user habits and apply them appropriately. They include temporary exceptions, changes in habits, and suggestions that users have not accepted. Sampled tasks undergo independent human review. We evaluate eight memory methods with Qwen3-235B-A22B and GLM-5.3-Flash. The best-performing memory method achieves 33.61\% macro accuracy across the four domains. Performance remains limited even when annotated evidence is supplied directly. These results expose a persistent gap between access to past interactions and the ability to infer and appropriately use user habits.

★ Beyond Factual Accuracy: A Benchmark for Evaluating Perspective Alignment in LLMs

Sunghoon Kim, Sangyeop Kim — Tue, 20 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Large Language Models often generate factually correct responses that fail to maintain user-requested perspectives, blending multiple perspectives and forcing users to filter through unwanted information---a perspective control failure that factual accuracy metrics cannot detect. We introduce PerspectiveBench, a benchmark quantifying perspective alignment by constructing scenarios where multiple valid reasoning paths coexist, and propose the Perspective Alignment Score (PAS) to measure adherence to specified perspectives. Experiments reveal that state-of-the-art models successfully include target content but fail to exclude non-requested perspectives. Providing documents through RAG improves factual grounding but does not resolve perspective drift, as models prioritize comprehensive coverage over strict user constraints. Code and data are available at https://github.com/sunghoon014/PerspectiveBench.

★ BrandEval: A Two-Track Multi-Agent Benchmark for Risk-Sensitive LLM Crisis Communication

Haotian Gan, Weidong Tang, Hanbin Sun, Yi Luo, Xintian Li, Jiayong Wang — Tue, 20 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Evaluating LLMs for crisis communication requires more than judging isolated responses, because public feedback and competitor actions can reshape trust over time. Existing benchmarks remain largely static and single-turn, offering limited evidence about long-horizon risk-sensitive communication. We introduce \textbf{BrandEval}, a two-track benchmark for product-crisis communication, focusing on public brand responses to product-related safety, quality, service, and trust incidents. BrandEval-Static evaluates stakeholder-aware strategic analysis, while BrandEval-Dynamic stress-tests response policies in \textbf{BrandPolis}, a controlled multi-agent sandbox with consumer feedback, competitor responses, and trust-risk trajectories. We also construct \textbf{BrandSRD}, a Chinese crisis-response dataset, and instantiate a reference Strategic Agent using \textbf{Strategic Rationale (SR)}. Across proprietary and open-source models, we find that static strategic quality is informative but incomplete: it aligns with some end-state outcomes but does not reliably predict volatility and delayed recovery; dynamic stress tests further expose tail-risk failures under manipulative or weakly grounded policies. The code and dataset will be available at https://github.com/htgan-ai/BrandEval.

★ When Empathy Misses the Goal: A Benchmark for Goal Displacement in LLM Advice

Dean Ariel, Guy Laban — Tue, 20 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Conversational large language models can respond with warmth and reassurance while failing to advance the user's stated goal. We conceptualize this failure as $\textbf{goal displacement}$, the substitution of persona-maintenance content for goal-relevant substance. To measure it, we introduce $\textbf{GDB}$ (Goal Displacement Benchmark), comprising 230 scenarios across nine domains and $4{,}134$ human-scored responses from 18 models. A $2 \times 2$ taxonomy crosses hard constraints with emotional distress, distinguishing failures to confront infeasible goals or combine empathy with practical guidance from reverse displacement, where unsolicited problem-solving replaces requested emotional support. A deletion-test rubric assesses whether substantive help remains after purely supportive sentences are removed. Independent raters achieve weighted $\kappa = 0.82$ against author scores for displacement severity. Trap displacement ranges from $0.0\%$ to $37.7\%$ and is strongly negatively correlated with Chatbot Arena Elo ($\rho = -0.706$, $n = 15$). Reverse displacement reaches $41.9\%$, including substantial rates in models with negligible trap displacement. A factorial probe supports a distinction from social sycophancy. Human-scored responses in the high-complexity, low-pressure condition exhibit $35.0\%$ sycophancy but $7.5\%$ displacement, with little association between the constructs. Matched OLMo-2 checkpoints show that mixed empathy-with-constraints displacement remains at $52.3\%$ after supervised fine-tuning and falls to $13.8\%$ after direct preference optimization. GDB provides a reproducible diagnostic of whether responses advance the operationalized user goal, supporting model comparison, post-training evaluation, and regression testing. We release the dataset, annotations, and rubric under CC-BY-4.0 and the evaluation code under MIT.

★ LEAPS: Long-context Evaluation of Adaptive Pattern Learning and State Tracking

Howard Yen, Nitya Mani, Alex Renda, Christopher S. Hardin — Tue, 20 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Long-context language models (LCLMs) power agents that act over long trajectories, from improving a codebase to running research experiments to forecasters that condition on lengthy histories. These applications draw on three distinct long-context capabilities: recall, where the answer is a fact present in the context; state tracking, where the outcome can be computed by performing a series of transformations to a specified state; and pattern learning, where the goal is to make a prediction by sampling from a distribution inferred from many observations. Existing benchmarks primarily focus on the first capability. In this work, we introduce LEAPS (Long-context Evaluation of Adaptive Pattern learning and State tracking), a collection of 12 novel datasets, spanning agentic workflows, behavioral modeling, and forecasting of market and world events. By construction, LEAPS is resistant to data contamination, extends to millions of input tokens, and uses reliable metrics. We evaluate 30 API and open models from 8K to 512K input tokens. LEAPS distinguishes models that perform near perfectly on recall tasks, finding that such models' state tracking and pattern learning abilities can vary significantly and degrade much more dramatically with increasing input lengths--even for pattern learning, where more observations should make learning a theoretically easier task. Moreover, thinking models often underperform their base counterparts, exhausting their generation budget or oversimplifying the task. We hope LEAPS can guide the development of the next generation of LCLMs.

★ UtilityBench: Measuring Language Model Utility When Errors Have Different Costs

Qifeng Lei, Zhiyong Yang, Zitai Wang, Shilong Bao, Linchao Zhang, Qianqian Xu — Tue, 20 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

As language models are increasingly deployed in real-world scenarios with important outcomes, the impact of their mistakes often extends far beyond what correctness can capture. In domains like medicine, a single error can lead to serious consequences and put patient safety at risk. However, mainstream evaluation protocols still treat all errors as equal, overlooking the spectrum of real-world costs. To address this, we propose a utility-based evaluation paradigm that explicitly incorporates the varying consequences of errors. Guided by this principle, we develop UtilityBench, a large-scale benchmark of 15,000 instances with verifiable answers, quantifying error severity using Best-Worst Scaling to elicit meaningful, data-driven cost estimates. These quantified utilities shape when a model should answer or abstain, guiding it to make decisions that align with the true importance of each scenario. As another merit, we introduce UtilityAUC, a metric that measures the area under the utility curve determined by a model's response ordering, thereby enabling utility preference alignment beyond error frequency. Our theoretical analysis demonstrates that UtilityAUC can distinguish between models where standard accuracy and AURC fall short. Empirically, evaluating 20 state-of-the-art models, we find that greater accuracy or more advanced architectures do not reliably translate to higher UtilityAUC, highlighting a gap between problem-solving ability and responsible decision-making in language model evaluation.

★ CXBench: A Benchmark for Process-Compliant LLM Agents in Regulated Customer-Experience Workflows

Boxi Yu, Yuzhong Zhang, Zhenfeng Su, Linfu Xi, Yinghan Yao, Shitao Chen — Tue, 20 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Agents 96%

Tool-using language agents increasingly perform regulated customer-service workflows, where a correct final state is insufficient: agents must also provide disclosures, obtain consent, and follow required procedures. Existing tool-agent benchmarks largely score final business states, accepting trajectories that reach correct outcomes through non-compliant processes. We introduce cxbench, a benchmark that evaluates outcome and process compliance separately across six regulated domains. Each task combines a customer request, a tool-based environment, and explicit process requirements. Rule cards link cited regulatory obligations to observable trace evidence. A hybrid grader combines deterministic checks for ordering and authority with an LLM judge for semantic requirements. We define a false-pass as an outcome-pass run that fails process checks. Across seven models and 120 tasks, 55.5% of the 840 retained executions are false-passes, representing 80.1% of outcome-passing runs; only 13.8% pass both checks. Process-inclusive evaluation changes the observed model ordering and exposes differences hidden by outcome-only scoring. cxbench makes these failures inspectable through trace-linked criteria.

★ AdaExploreBench: Benchmarking LLM Agents for Proactive Exploration and Test-time Adaptation in Evolving Environments

Rui Wang, Zishen Lai, Jiayu Liu, Cheng Qian, Jeonghwan Kim, Heng Wang — Tue, 20 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Large Language Model (LLM) agents are increasingly deployed in unfamiliar and evolving environments, requiring them to proactively explore, understand, and adapt to environmental dynamics. However, existing benchmarks rarely test whether agents can proactively discover how an environment works and adapt as it evolves. To address this gap, we introduce AdaExploreBench, a benchmark for proactive exploration and test-time adaptation in evolving environments. By instantiating environmental dynamics as hidden rules, AdaExploreBench requires agents to actively acquire evidence and infer the rules. Under implicit rule drift, the latent rule changes without announcement, requiring agents to detect from degraded performance that prior knowledge is stale and re-explore. Experiments on eight leading LLMs show that proactive and adaptive exploration remain challenging: GPT-6 Sol achieves only 61.9% calibrated accuracy, while providing sufficient evidence or removing rule drift improves average performance significantly by close to 10%. These results reveal a critical bottleneck in current LLM agents: they can often reason effectively from informative evidence once it is available, but struggle to actively acquire the right evidence and recognize when previously acquired evidence is no longer valid.

★ Scaling Synthetic Pre-training Data and Evaluation for Indic LLMs

Tue, 20 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, New model 58%

Building capable language models for India requires scalable multilingual data and robust evaluation. We introduce Nemotron-Pretraining-Indic-v1, a synthetic pretraining corpus of approximately 2.83 trillion tokens covering all 22 scheduled Indian languages in native script and twelve in romanized form. It is the largest openly released Indic pretraining corpus. Using our GPU-accelerated pipeline, we translate curated English documents, assess translation quality, and repair low-quality outputs while preserving technical content and document structure and maintaining source traceability. Through continued-pretraining experiments across twelve languages, we show that combining our synthetic corpus with natural Indic text improves knowledge and reasoning over a public-corpus baseline. Additional experiments show how data volume, script coverage, and Indic training share affect English performance. We also introduce the Indic-LLM-Eval-Suite, a multilingual benchmark suite for evaluating Indic language models across chat, instruction following, reasoning, knowledge, code, and translation. The suite combines existing Indic benchmarks with task-aware translations of English benchmarks, using automated quality assessment, targeted correction, and native-speaker review in a human-in-the-loop pipeline. We use the suite to comprehensively benchmark contemporary open-weight Indic and multilingual models across capabilities and languages, revealing substantial variation across tasks and languages. Together, these resources provide a scalable foundation for developing and evaluating more capable Indic language models.

★ DeepSEA: A Multilingual Evaluation Benchmark for Southeast Asia

Jiuzhou Han, Abhishek Bijay Mishra, Trevor Cohn, Shubham Mittal — Tue, 20 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

While language models have demonstrated remarkable capabilities, their performance remains disproportionately skewed towards high-resource languages. This disparity is particularly acute for the 700 million inhabitants of Southeast Asia, where vast linguistic diversity remains under-represented in existing benchmarks, often suffering from limited language coverage or narrow task scopes. To address this, we introduce DeepSEA, a comprehensive benchmark constructed from established datasets through professional human translation into 12 Southeast Asian languages, spanning both high-resource (e.g., Indonesian, Thai, Vietnamese) and low-resource (e.g., Burmese, Khmer, Lao) tiers across five distinct tasks: mathematical reasoning, scientific reasoning, question answering, abstractive summarization, and machine translation. Extensive evaluation of frontier proprietary and open-weights models reveals that while frontier models like Gemini-3.7-Flash maintain robustness across resource tiers, smaller models face severe performance degradation on low-resource languages. Crucially, models perform better in reasoning and summarization when pivoted through English rather than target languages, exposing a critical dependency on English representations, especially for low-resource languages. Beyond these observations, we provide a cross-lingual error analysis and a cross-script tokenizer fertility analysis. We hope these findings shed light on the development of next-generation inclusive multilingual models.

★ MoveBench: A Rubric-Based Benchmark for Evaluating Large Language Models in Sports and Health Coaching

Yitong Li, Jiang Bian, Jingling Wang, Pan Ding, Chengying Gu — Tue, 20 Oct 2026 00:00:00 +0000

99% · Evaluation 99%

Despite rapid advances in large language models (LLMs), their deployment in the sports and health domain remains high-risk due to the absence of domain-specific evaluation benchmarks. Existing general-purpose benchmarks (e.g., MMLU and HELM) primarily assess broad knowledge and reasoning capabilities, but fail to capture critical domain requirements such as exercise safety, physiological appropriateness, personalization, and actionable coaching quality. This gap is particularly concerning in fitness and health applications, where incorrect or unsafe recommendations can lead to physical harm. To address this limitation, we introduce MoveBench, the first comprehensive benchmarking framework specifically designed for evaluating LLMs in sports and health coaching. The proposed benchmark integrates heterogeneous task formats, including multiple-choice questions (MCQs) and open-ended question answering (OpenQA), grounded in real-world user scenarios and paired with structured rubrics emphasizing safety compliance, scientific validity, and training logic. Through large-scale evaluation of 15 diverse LLMs on 809 questions with 5,012 rubric-level criteria, our results reveal substantial performance variability, particularly in safety awareness, structured program design, and personalization capability, highlighting a critical gap between knowing and doing. Our work establishes a foundational evaluation standard for sports and health LLM applications and provides a reproducible framework for advancing safe, reliable, and context-aware AI-driven coaching systems. Open resources are available at https://github.com/SatoSakula/MoveBench.

★ KEEP-Bench: Evaluating Evidence Preservation in Context Compression

Ziqing Man, Di Wu, Jia-Chen Gu, Junyi Zhang, Kai-Wei Chang, Nanyun Peng — Tue, 20 Oct 2026 00:00:00 +0000

99% · Evaluation 99%, Hallucination 49%

Context compression improves long-context efficiency but may introduce faithfulness issues through an information bottleneck. Existing evaluations rely on QA metrics, but a reader can still answer correctly from an unfaithful compression, so these metrics underestimate such loss. To bridge this gap, we introduce KEEP-Bench, a diagnostic benchmark that directly assesses compression faithfulness with 4,102 cases across three challenging settings that require keeping evidence for plausible reference resolutions, explicit qualifiers, and minority perspectives. Experiments show that even with tailored prompting, the strongest models remain unreliable: despite context windows exceeding 100K tokens and inputs averaging only 3.1K tokens, the best open-weight model (31B) preserves all required evidence in only 66.8% of cases, and the best proprietary model in only 61.6%. Performance also varies sharply across evidence structures, so conclusions from limited structures may be biased. We further introduce a metric, Evidence Preservation Return (EPR), to relate evidence recall to inference FLOPs, and find that it peaks at moderate scales (4B-8B), where recall remains low. Additional output budget mainly benefits already strong models and widens capability gaps, yet the stronger preservation of these models comes at a steep computational cost. These findings highlight the importance for future work to investigate compressors with strong information preservation under a small computation budget, rather than relying on a large output budget or model scaling.

★ OR-License: Evaluating LLM Agents as Real-World Operations Research Practitioners

Peisong Wang, Feixue Liu, Yize Meng, Bowen Liu, Ruoyu Chen, Gaoyang Dai — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 67%, Hallucination 36%

Can large language model (LLM) agents deliver expert-level operations research (OR) solutions directly from original client materials, without prior expert reformulation? Existing OR benchmarks typically begin with self-contained descriptions or predefined optimization tasks, underrepresenting the translation of original business materials into professionally validated solutions. We introduce OR-License, an end-to-end benchmark of 31 instances across 10 industrial scenarios from paid OR engagements totaling USD 3.68 million. It preserves original requirements, input data, deliverable formats, and information boundaries after de-identification. To study information acquisition, we introduce ORSim, a controlled client simulator that discloses business facts without supplying OR expertise. Hidden evaluation checks feasibility and compares objective quality against original expert solutions accepted by clients before benchmark construction. Deployment-Adjusted Quality (DAQ) averages expert-relative quality across instances, assigning zero to infeasible deliveries. Across 11 frontier LLMs using coding-agent harnesses, several exceed the expert reference in conditional objective quality, but none matches its DAQ; the strongest primary Claude Code configuration achieves 88.39% strict feasibility and a DAQ of 0.87 against the expert reference of 1. Client information acquisition extends beyond clarification into solution development and validation, with some later stages showing higher confirmed issue-resolution proportions. Trajectory analysis identifies formulation and implementation failures, while task-equalized research-chain completion rates range from 24% to 45% across models on evaluable tasks. These findings motivate sustained information-gap resolution, faithful problem formulation, and evidence-linked self-improvement for reliable long-horizon agents.

★ Agents' Last Doubt: Knowing When to Act and When to Step Back in Tool-Using Agents

Qisai Liu, Qingyue Jiao, Bangwei Guo, Wenxi Dai, Changyi Yang, Jiahe Liu — Tue, 20 Oct 2026 00:00:00 +0000

98% · Agents 98%, Evaluation 80%, New model 74%

Tool-using LLM agents repeatedly face decision points under uncertainty, choosing between autonomous action, additional information gathering, and human escalation. These decisions require balancing task success against interaction cost and the risk of irreversible errors. Learning to make these trade-offs is challenging because existing interaction trajectories record only the outcomes of actions taken, leaving the consequences of alternative actions unobserved. To address this gap, we construct a counterfactual dataset containing 3,115 action outcomes from 1,036 decision points in ToolSandbox. At each decision point, we execute each legal action and evaluate its downstream outcomes based on task completion, interaction cost, and irreversible errors. Using this dataset, we train SURE-Act, a controller that selects actions for a frozen backbone agent at inference time. Our experiments show that learning to rank alternative actions improves offline action-selection utility by 0.0309 over a regression-only baseline. We further show that adding SURE-Act enables end-to-end gains of up to 0.0575 in ToolSandbox's Overall Score with the Qwen3-4B agent. Moreover, SURE-Act outperforms adapted decision rules from SAGE-Agent and Calibrate-Then-Act by 0.1506 and 0.1438, respectively. This work establishes the first counterfactual dataset for ToolSandbox and provides actionable insights into decision-making under uncertainty for tool-using agents.

★ AM-Bench: A Unified Taxonomy and Evaluation Suite for Agentic Misalignment

Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 38%

As LLM-based agents are deployed with increasing autonomy in real-world environments, a critical failure mode emerges: agentic misalignment, where agents choose harmful actions over failure. However, existing safety benchmarks primarily evaluate whether agents comply with explicitly harmful instructions, leaving the conditions that cause agents to choose harmful actions poorly understood. To address this limitation, we propose AM-Bench, a unified taxonomy and 35 scenario evaluation benchmark spanning seven industries for agentic misalignment grounded in the fraud triangle, a well-established criminological framework explaining goal-directed misconduct. We adapt this framework to agentic environments and construct our evaluation suite and detection systems along three dimensions: perceived threat, perceived opportunity, and reasoning. We operationalize each dimension into structured subcategories rooted in prior theory: five threat categories derived from the instrumental convergence thesis, seven opportunity categories drawn from the Emergent Strategic Reasoning Risks (ESRR) taxonomy, and a novel four-question binary reasoning framework synthesized from prior literature. Our evaluation of 10 frontier models reveals a 62-percentage-point gap in misalignment across scenarios, with Gemini 3 Flash and Gemini 3.1 Pro exhibiting the highest levels of agentic misalignment while Claude Sonnet 4.6 displays near-zero rates across all conditions. Our reasoning framework further reveals that models may exhibit harmful behaviours without explicit situational awareness or explanation, complicating both mitigation and detection. AM-Bench scenarios and code are available at https://github.com/AI4Collaboration/Agentic-Misalignment-Bench

★ BanglaShield: A Script-Robust Safety Classifier and Benchmark for Bangla Native, Romanized, and Code-Mixed Prompts

Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, New model 58%

Safety classifiers for large language models (LLMs) are typically built and evaluated on standard-script text, yet Bangla, like many lower-resource languages, is routinely written in native script, in Latin-script romanization, and code-mixed with English, Hindi, or Urdu. We introduce BanglaShield, a resource and model for graded prompt-safety classification across these written forms. We consolidate 10,245 Bangla prompts from six sources under a four-level taxonomy (Safe, Offensive, Harmful, Severe) and render each prompt in three aligned forms that share labels and a group-aware split, enabling strictly paired comparisons. Benchmarking six contemporary LLMs, including a Bangla-specialized one, on 36,882 classification queries and 24,318 judged responses, we find that LLM classifiers label 47.0% of unsafe native-script prompts as Safe, rising to 63.4% under romanization, while response-level safety accuracy drops by 10.3 points. Three multilingual guard models also pass 30.3-57.3% of unsafe native-script prompts. Using a model's own classification as an input filter reduces harmful compliance by only about five points. In contrast, BanglaShield, a compact fine-tuned encoder, reaches 87.8 macro-F1 on native-script and 84.4 on romanized prompts, outperforming the best LLM on each form by 42.8-46.1 points with ~130x fewer parameters than Gemma 4 31B. We release the corpus, splits, judgments, and model.

★ EdgeKernelBench: Unified LLM-Driven Kernel Generation for Edge Backends

Yuhua Zheng, Xueyi Chen, Xinze Zou, Xin Li, Haoji Hu, Huan Wang — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Recent works show that LLMs can generate kernels with strong performance on server GPUs. Edge deployment presents distinct challenges because the hardware is heterogeneous, the software stacks are fragmented, and the resource budgets are tight. LLM kernel generation across diverse edge backends has yet to be studied systematically. We introduce EdgeKernelBench, which combines a unified IR deployment framework with a hierarchical benchmark for edge workloads. The unified IR captures reusable computation, scheduling, and memory decisions and connects to each target backend through a compiler interface. The benchmark spans progressively larger program scopes, from individual operators to compact model graphs, across representative edge application domains. We instantiate the design with TVM as a reference compiler and validate it across heterogeneous edge backends. With a single generation attempt, the best model achieves 64.5% Correct@1 but only 25.5% Fast@1.0. This gap shows that functional correctness does not consistently translate into performance gains. We therefore introduce the Layered Edge Kernel Agent (LEKA), which organizes optimization through a layered IR. On the agent evaluation subset, LEKA achieves 43.5% Fast@1.0, compared with 13.0% for direct generation and 26.1% for Claude Code. LEKA and Claude Code use the same base model and optimization budget. Together, EdgeKernelBench and LEKA support unified evaluation and layered optimization of kernels generated by LLMs across edge backends.

★ fxBench: Evaluating and Understanding Formula Suggestions in Spreadsheets

Sanket Mhatre, Sumit Gulwani, Vu Le, Yasharth Bajpai, Gust Verbruggen — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Efficiency 36%

Spreadsheets are among the most widely used tools for computations, and formulas are their core building blocks. Besides agentic spreadsheet tools, proactive formula suggestions enable significant productivity gains and are now available in popular spreadsheet software (Microsoft Excel and Google Sheets). Such suggestions pose interesting challenges: the right context must be collected and represented for a machine learning model, which must then be able to write a correct formula, in a way that is fast and cheap enough to trigger on every = (or subsequent keystrokes for auto-completion) without taking users out of their editing flow. Surprisingly, there is no public benchmark for formula suggestions on real spreadsheets. In this paper, we therefore (1) introduce fxBench as a manually curated benchmark of 503 formula suggestion tasks based on Sheetpedia, (2) break down the formula suggestion problem into four capabilities (gather context, understand context, understand intent, write formula) that a formula suggestion system requires, and (3) evaluate existing and new approaches to understand bottlenecks in these capabilities that pose interesting research directions. Specifically, we annotate each benchmark with the cells that describe intent to determine the theoretical performance limit of a given context selection strategy, we evaluate performance of small and large models on different representations of context, including a dense format with high theoretical coverage at low token counts, and we fine-tune smaller language models to show that we can learn to understand such dense formats. Additionally, we perform quantitative and qualitative analysis on failure modes to understand bottlenecks in the formula writing capability; the best configuration only achieved 58.8% correct suggestions.

★ Can LLM Agents Resolve Service Failures? Simulating Multi-Turn Complaint Handling under Information Asymmetry

Wailing Ng, Mengze Hong, Chen Jason Zhang, Shiqi Fan — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, New model 37%

LLM agents are widely adopted in service chatbots for rule-guided support and information access, yet emotionally demanding complaint handling still relies heavily on human agents. We introduce ServiceBench, to our knowledge the first benchmark dedicated to complaint handling, with 2,500 scenarios spanning 184 complaint categories and five service industries. We propose Asymmetric Progressive Complaint Simulation (APCS) to simulate realistic complaint handling, together with a justice-based evaluation framework. Our experiments reveal that even capable LLM agents struggle to handle complaints effectively. They lack the ability to proactively gather information and strategically guide conversations toward resolution, often resorting to premature compensation. We further show that targeted post-training with SFT and DPO substantially improves these capabilities, enabling a smaller model to match or surpass much larger models on several outcomes.

★ BanglaBBQ: A Bias Benchmark for Question Answering in the Bangladeshi Context

Sumit Anthony Gomes, Yasin Rahman, Taeeba Tasnia, Faiza Nooren Suhita — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

This paper introduces BanglaBBQ, a question-answering bias benchmark designed to capture stereotypes grounded in the Bangladeshi context, in both Bangla and English. Following KoBBQ and PakBBQ, BanglaBBQ adapts the original Bias Benchmark for Question Answering (BBQ) by classifying its templates into three groups: directly translated (DT), modified to align target groups or contexts with Bangladesh (TM), and removed when deemed unsuitable for the Bangladeshi context (SR). This process retains 140 templates across six categories, comprising 76 DT and 64 TM templates. Five newly authored (NA) categories, based on stereotypes common in Bangladeshi society, are then added: marital status, age and political perception, regional origin, educational background, and mental-health status. Native Bangladeshi annotators judged all 466 newly authored templates, and 372 were retained because a strict majority of English annotators identified the intended target and rated the attitude as common enough. Each newly authored template is linked to a documented source and a verbatim quote from it. Each evaluated template also has a counter-stereotypical version of its disambiguated context. The dataset was evaluated using two open-weight models (Qwen3-4B and Qwen3-8B) and two closed models (GPT-4.1-mini and Gemini 2.5 Flash). All four lean towards the stereotyped person when the context is ambiguous, in both languages, and the effect of language differs by model, so English results do not predict behaviour in Bangla. On the adapted BBQ templates, both Qwen3 models are also less accurate when the context contradicts a stereotype than when it confirms it. On the newly authored templates, the closed models follow counter-stereotypical evidence almost perfectly, and all four models are more biased on the templates that annotators validated than on those they rejected.

★ Medical Fine-Tuning Reshapes, Not Reduces, Patient-Facing AI Risk: A Rubric-Based, LLM-Judge Evaluation Framework

Savan Doshi — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 69%

Patient-facing medical question answering with large language models (LLMs) is moving from research demonstrations to production deployments, yet evaluation remains anchored on factual correctness. Factuality is necessary but not sufficient: a response can be factually grounded and still issue a directive that, if acted on, would harm the patient. We introduce a risk-sensitive evaluation framework that scores responses by the presence and severity of six categories of risk-bearing language (treatment directives, contraindications, dosage expressions, urgency-triage cues, high-alert medication mentions, and overconfident assertions), anchored on published clinical harm taxonomies. Severity is classified by a three-judge LLM ensemble. We apply this framework to six 7-9B instruction-tuned models, three general (Mistral-7B, Phi-3.5-mini, Llama-3.1-8B) and three medically adapted (BioMistral-7B, OpenBioLLM-Llama3-8B, Med42-Llama3-8B), on 200 real patient questions from the K-QA benchmark. Two findings emerge. First, ensemble risk-bearing language is statistically independent of factual hallucination rate (pooled Spearman $\rho = -0.065$, $n = 300$, 95% CI $[-0.18, +0.05]$; distance correlation $= 0.10$, independence test $p = 0.24$). Second, medical fine-tuning does not reduce aggregate risk but reshapes it: continued-pretraining adaptations (BioMistral, OpenBioLLM) consistently suppress high-alert medication mentions while amplifying overconfident assertions, whereas clinical instruction-tuning (Med42) suppresses dosing only, without amplifying overconfidence. Per-category Krippendorff $\alpha$ is fair to good for the four linguistically anchored categories (0.53-0.71) and low for the two clinical-judgment-dependent categories (0.19, 0.31), indicating where automatic safety evaluation is and is not yet reliable. We release the rubric, judge prompts, and evaluation code.

★ Pattern-Driven Evaluation of LLM Code Optimization: Isolating Compiler-Transcendent Semantic Reasoning

Wesley Lu, Runzhou Liu, Hongru Du, Xinyao Yi — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Existing evaluations of LLM-based code optimization primarily measure end-to-end speedup, obscuring whether a model actually performs the intended optimization. For compiled languages, this is especially problematic: observed gains may reflect transformations already recoverable by modern optimizing compilers rather than optimization behavior attributable to the model itself. We present PatternBench, a benchmark for studying compiler-transcendent optimization, i.e., performance improvements that remain after aggressive compiler optimization. PatternBench targets residual optimization opportunities across seven classes of transformations and pairs runtime evaluation with an AST-based optimization faithfulness analysis that verifies whether the intended optimization is realized in the generated program. Experiments with frontier, open-weight, and fine-tuned models show that runtime speedup alone provides an incomplete account of optimization capability. Optimization faithfulness is strongly associated with realized speedup, and fine-tuning improves both faithfulness and optimization effectiveness. These results suggest that evaluation of LLM-based code optimization should measure not only performance outcomes, but also whether the target transformation was successfully executed.

★ MisleadQA: Diagnosing LLM Failures under Misleading Context

Hui Wang, Bohao Chu, Qianli Wang, Shuning Zhang, Norbert Fuhr — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 57%

Large language models (LLMs) are widely used to answer questions from retrieved or user-provided context. However, provided context can be factually true yet out-of-scope, describing a different entity, time, or location from the one the question targets, and still lead models to a wrong answer. Prior work has mainly studied factually false or irrelevant context. It remains unclear whether such out-of-scope context misleads models as much as false context, and whether adding gold context prevents models from being misled by either. We introduce **MisleadQA**, a diagnostic dataset of multiple-choice questions with an abstention option. Each question pairs gold context (G) with true but out-of-scope context (S) and false but in-scope context (F) supporting the same wrong answer. We evaluate six open-weight LLMs. On questions they answer correctly without context, providing S alone leads them to switch to this wrong answer in 29.3% of cases, and providing F alone does so in 75.4%. Adding G nearly removes the effect of S, but F still misleads models in 21.0% of cases, especially when F comes first. A context-faithful instruction reduces these errors mainly through abstention and lowers accuracy even with G. These findings suggest that evaluations should check whether context applies to the question, not only whether it is true, and should report correct, misled, and abstention rates together. Data and code are available at [https://anonymous.4open.science/r/MisleadQA](https://anonymous.4open.science/r/MisleadQA).

★ Toward Agentic Retrieval-Augmented Generation for Enterprise Retail Workflows over Heterogeneous Data Sources

Jeonghwan Choi, Jibin Hwang, Hayoon Park, Hyeonjae Cheon, Hwanjun Song — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 63%, Hallucination 48%

Enterprise retail workflows require agents to chain evidence across heterogeneous sources, from structured product and order records to unstructured inquiries and reviews, yet existing agentic RAG benchmarks rarely capture this challenge. To enable controlled evaluation of such workflows, we introduce RetailBridge, a compositional retail benchmark built from persona-driven simulations. Its questions are constructed by linking entities and intermediate answers across heterogeneous sources, requiring agents to identify the relevant sources, carry evidence between them, and produce a grounded answer. To systematically evaluate LLMs on these retail workflows, we develop RetailRAG, a retail-aware framework equipped with auto-schema briefing, source-routed plan chaining, and diagnostic plan revision. RetailRAG consistently outperforms existing agentic RAG baselines across all evaluation axes, serving as a strong reference system for assessing LLM agents on retail workflows. On RetailBridge, evaluating ten LLM backbones with RetailRAG reveals multi-source chaining as the dominant bottleneck across providers and model scales, suggesting that reliable retail agents require explicit cross-source orchestration rather than scaling alone. Code and data are publicly available at \url{https://github.com/DISL-Lab/RetailBridge}.

★ Explore or Confirm? LLM Agents Misallocate Measurements Under Noise

Abdul Karim, Matt Keon — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

As model reasoning gets cheaper, the physical experiment becomes the slow and costly step in discovery. Each experiment returns a noisy reading: a high value can mean a good candidate or a lucky draw. With a fixed budget, an AI agent must keep choosing between exploring a new candidate and confirming a promising one by re-testing it. We measure that choice in Wellocate, a simulated assay campaign with known true values. It is built on a generated landscape and on two real protein datasets, GB1 and PhoQ, scrambled so no model can answer from memory. We test five language models as minimal agents, with no tools and no memory beyond the history in each prompt. They misallocate in both directions. Three over-confirm, spending most of the budget re-testing. On the generated landscape one under-confirms, trusting single readings, and there every model ends with a worse final pick than a basic scripted strategy. Each failure has its own cause. The over-confirmers run short of new candidates worth trying: a list of untried candidates cuts their re-testing and, on real protein data, makes two of them pick much better. A running summary of their readings or a description of the landscape changes it by at most five percentage points. For the under-confirmer, re-testing its top candidates in a final round, replayed on its own campaigns, recovers only about a quarter of what it loses by picking the wrong candidate. This suggests confirmation has to run through the whole campaign. We release Wellocate as a benchmark and reinforcement-learning environment with frozen splits and scripted baselines. With it, anyone can measure how an agent splits its budget before trusting it with real experiments, and train agents to split it better. Code, data and the environment are available at https://github.com/Xstendyl/wellocate.

★ MemBubble: Benchmarking How Agent Memory Shapes LLM Response Diversity

Jiajun Tan, Fei Sun, Xinyu Ma, Shixuan Zhang, Du Su, Qi Cao — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Memory serves as the core component of LLM-based agents, enabling personalized responses via extracting and using user preferences from past interactions. However, its use may reduce response diversity, thereby narrowing the range of information available to users. Existing evaluations rely largely on synthetic interactions and cover few memory systems, leaving the effects of memory on response diversity insufficiently evaluated in real-world scenarios. We introduce MemBubble, a benchmark of 1000 instances built on preferences extracted from real user-assistant conversations. To measure the extent to which memory concentrates response distributions, we construct test queries in scenarios distinct from user preferences while preserving topical relevance through shared preference categories. Across multiple LLMs and memory mechanisms, entropy decreases by 34.7% on average compared with directly providing the full interaction history, highlighting the important role of memory processing in diversity loss. We also explore simple user-side interventions and changes in information availability but find that these do not effectively close the diversity gap. Our findings highlight the need to consider preserving diversity and user access to information alongside improvements in personalization when designing and using agent memory systems.

★ A Thousand Faces: A Persuasion Benchmark with Experience-Grounded Persuadees Informed by Psychology and Pragmatics

Ziqing Zhuang, Dingyi Zhang, Deyu Zhou — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

People can express the same beliefs and desires yet need different reasons to change their minds. What makes an argument relevant depends on how those views arose: experiences shape what a person finds credible, what matters to them, and what could prompt reconsideration. These personal grounds may also be expressed indirectly, through contextual clues in conversation. Benchmarks that specify beliefs and desires without these connections can reward arguments that match a description while overlooking whether they address the person's reasons. We introduce A Thousand Faces (ATF-Persu), with 1,000 personas across 200 topics and nine domains. Drawing on personality and attitude theories and conversational implicature, we construct five personas per factual context, linking experiences to judgments, expression, and grounds for revision. Evaluations show lower persuasion success than three comparison benchmarks under their respective protocols. Complete backgrounds yield broader use of supplied personal grounds and higher persona faithfulness than simplified specifications; blinded reviewers also rate ATF-Persu above CToMPersu on identifiable personal reasons, their influence on responses, and contextual expression. The dataset is publicly available at https://huggingface.co/datasets/Slandes/ATF-Persu.

★ Where and Why Search-Augmented LLMs Fail When Reasoning over Rapidly Evolving Knowledge

Tianyi Hu, Bhavyajeet Singh, Niket Tandon, Akhil Arora — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 79%, Hallucination 40%

As the world changes, new information is published while older records remain online. This complicates reasoning for large language models (LLMs), which must retrieve evidence and resolve conflicts to determine which information remains valid. Existing evaluations treat information seeking as a black box, revealing little about where and why LLMs fail. To address this gap, we introduce TREK, a suite of three components: TREKTRACE is a plug-and-play framework that attributes failures to specific stages of retrieval and reasoning. We apply it to TREKBENCH-WILD, a meta-benchmark of real-world questions that require complex reasoning over rapidly evolving knowledge. Surprisingly, the main bottleneck is post-retrieval reasoning rather than retrieval: frontier models are often misled by conflicting evidence even after retrieving all the information needed to answer. To study this bottleneck, we build TREKBENCH-CTRL, a controlled benchmark with distractors spanning temporality, relevance, and trustworthiness, and show that benchmarks dominated by outdated distractors substantially underestimate task difficulty. Together, TREK provides a testbed for diagnosing and improving reasoning over rapidly evolving knowledge.

★ Aligning Cultural Behaviors in Arabic Language Models via Preferences

Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, New model 33%

Arabic LLMs have improved rapidly, but aligning them with country-specific cultural behavioral preferences remains underexplored. Previous work has focused primarily on instruction-tuning for factual knowledge, overlooking nuanced cultural behaviors. We propose ArabPref, which to our knowledge is the first large-scale pan-Arab preference data set for the alignment of cultural behavior conditioned on the country. Our contribution includes two resources in English and Modern Standard Arabic: a culturally grounded preference dataset for training and evaluation, and a multiple-choice benchmark designed to assess culturally aligned behavior across nations. All test data are validated by native speakers to ensure authenticity. To optimize cultural alignment, we experiment with post-training methods across multilingual and Arabic-centric language models, evaluating performance on both generation tasks and our cultural multiple-choice benchmark. By releasing an evaluation benchmark, we provide a foundation for systematically studying and improving culturally aligned behavior in language models.

★ Towards Figurative Language Understanding in Moroccan Darija: An Evaluation of Large Language Models

Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Figurative language remains a major challenge for Machine Translation (MT) systems, as the intended meaning of such expressions often goes beyond literal translation. Although recent advances in Large Language Models (LLMs) improved the translation of Arabic dialects, benchmark resources for evaluating figurative language translation in Moroccan Darija remain scarce. To this end, we introduce the Moroccan Darija Figurative Language Corpus (MDFC), a benchmark dataset for MT and multiple-choice question (MCQ) selection covering idioms, metaphors, and proverbs. Using MDFC, we evaluate 19 open- and closed-weight language models and MT systems at both the expression and contextual levels under zero-shot and six-shot settings. We also conduct layer-wise similarity-based probing to determine how model representations distinguish figurative from literal meanings. Our findings reveal that closed-weight models achieve the strongest translation performance, while the benefits of contextual information vary across models and tasks. Differences between translation performance and MCQ accuracy highlight the distinction between generating an appropriate figurative translation and recognizing the intended meaning among candidate translations.

★ DiaLong: Benchmarking Reminiscence in Long-Context Dialogues

Owen Bianchi, Samer Aslan, Ryunosuke Saito — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

How well can transformer-based Large Language Models (LLMs) retain and recall memories in long contexts of dialogue? LLMs are known to forget long-range factual information in their context, yet the field lacks a comprehensive benchmark to quantitatively assess these shortcomings. We present DiaLong, a long-context dialogue dataset and accompanying memory task that standardizes evaluation in this space. It consists of 100 unique conversations created by humans and then augmented with LLMs, extending beyond lengths found in existing datasets. Our selected memory task is fact recall: classify 40 candidate facts about a dialogue as true or false. This task directly probes the model's ability to actively retain and retrieve information, serving as a critical measure of memory within extended conversational contexts. In testing both the latest LLMs and proposed memory enhancements on DiaLong, we demonstrate that more work needs to be done to enable LLMs with better memory performance. By enhancing memory capabilities, LLMs can improve in maintaining coherence, relevance, and human-like interaction over extended conversation. Improving these capabilities will help LLMs fulfill diverse user needs and expectations in various conversational scenarios.

★ Benchmark for Structured Retrieval-Augmented Generation in Perovskite Photovoltaics

Qiwei Han, Chi Zhou, Gaoxiang Yuan — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 49%, Hallucination 36%

While retrieval-augmented language models increasingly serve as literature assistants in the physical sciences, existing benchmarks evaluate them primarily on qualitative text overlap or through model-based scoring rather than against verified experimental ground truth. In experimental materials disciplines such as perovskite photovoltaics, however, the most critical inquiries are inherently quantitative and aggregate measurements across dozens or hundreds of papers. Here, we present a quantitative question-answering benchmark where all 755 ground-truth answers are computed directly from 45,441 measured device records (40,889 human-curated records through 2021 and 4,552 records extracted from 2022-2026 literature). Benchmarking six retrieval architectures over 158,274 passages from 25,662 papers using an open 14B model on a single 24 GB GPU, standard passage retrieval tops out at 16-23% accuracy (compared to 1.6-6.5% zero-shot). In contrast, equipping the language model with read-only SQL execution over the structured device table elevates accuracy to 51.5% on historical and 50.8% on prospective literature -- a 33.01.8 percentage-point advantage over hybrid text retrieval. Aggregate queries spanning extrema, distributions, and filtered record counts prove virtually unsolvable via text retrieval alone, demonstrating that systematic table extraction is an operational prerequisite for literature-scale quantitative inquiry.

★ Neither Over- nor Under-Constrained: Evaluating LLM Personalization under a Constraint Qualification Audit

Xuran Li, Guanqin Zhang, Imran Razzak, Hakim Hacid, Eleanna Kafeza, Hao Xue — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Accurate evaluation of LLM personalization is hindered by current evaluation paradigms that over- or under-constrain the response space, yielding frequent false rejections and false acceptances. To reduce these false rates, we introduce ConQual-Audit, an audit-based framework that evaluates LLM personalization through a three-stage process: (1) extracting semantic constraints from user profiles and task-grounded reference; (2) qualifying constraints by pruning unnecessary constraints and recovering missing constraints; and (3) verifying candidate responses against the audited constraints via natural language inference. Across PrefEval and LaMP, ConQual-Audit outperforms nine baselines, attaining top AUCs (0.88 and 0.90) and the lowest false rate (0.21) on the leanest constraint sets at an amortizable audit cost. Crucially, this constraint audit is portable: integrating it into ExPerT and PREF helps revise smaller, higher-quality constraint sets to reduce their evaluation false rates.

★ When to Stop Asking: Factor Coverage Bias and LLM Determinability Judgments with Underspecified Information in Medicine

Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Clinical decision-making often begins with underspecified information and requires not only gathering missing information but also knowing when to stop. This calls for judging determinability (whether the available information is sufficient for a decision) by considering whether missing information could change the conclusion, not simply the amount of available decision-relevant information. Reliance on this amount as a cue for determinability, which we term \textit{factor coverage bias}, could undermine the trustworthiness of large language models (LLMs) through premature conclusions or excessive information gathering, such as unnecessary invasive tests. Yet whether LLMs exhibit this bias remains insufficiently studied. To address this gap, we construct a controlled pairwise benchmark in which LLMs identify the determinable case within each determinable--undeterminable pair. We add a decision-relevant factor to either side while keeping each case's determinability unchanged, and similarly add irrelevant text to distinguish this bias from context-length effects. Adding a relevant factor made LLMs more likely to select the augmented case as determinable even when it remained undeterminable, an effect that persisted with increased reasoning effort, while adding irrelevant text made them less likely to select the augmented case, an effect that reasoning substantially reduced. These findings identify factor coverage bias as a failure mode distinct from context-length effects and a potential risk to LLM trustworthiness. Our work provides an evaluation framework for LLM trustworthiness under underspecification.

★ When Retrieval Partially Conflicts with Skills: Understanding Procedural Hallucination in Language Models

Guanqun Jiang, Pengfei Cao, Kang Liu, Jun Zhao — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 96%, Agents 39%

Language models increasingly use skills that provide reusable instructions and procedures for task execution. After a relevant skill has been selected, the model may need to apply its governing procedure together with retrieved task context. When the retrieved context is useful but contains cues that violate the skill, the model must retain compatible information while rejecting the conflicting cues. We study failures in which the model adopts such cues and produces a retrieval-supported but skill-invalid answer, which we call procedural hallucination. We introduce SourceClash, a matched benchmark with 1,271 bundles and 6,355 prompts across ten procedural domains. Each bundle fixes the task, skill specification, and answer space while varying the relation between the skill and retrieval. Across nine models, valid-answer rates average 25.7% under conflicting retrieval, compared with 78.2% and 72.8% under two supportive retrieval settings. Mechanistic analysis of open-weight models shows that the answer-preference difference typically reaches its maximum shortly after generation begins. Residual replacement moves the model toward valid answers, while attention decomposition attributes most of the repair to value vectors. Intervention at the identified value pathway also improves final answer validity. These results characterize how partly conflicting retrieval changes skill-guided generation.

★ FrontierRefactor: Benchmarking Multi-Scale Performance-Oriented Codebase Refactoring

Shijian Wang, Fanqing Meng, Jiajun Chen, Qiguang Chen, Linxin Song, Runhao Fu — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 48%

Autonomous codebase refactoring requires coding agents to restructure working software while strictly preserving externally observable behavior and prioritizing execution performance and compute efficiency over stylistic preference. Existing benchmarks: refactoring benchmarks assess functional correctness while overlooking runtime regressions, whereas code performance benchmarks target isolated micro-kernels that ignore codebase-scale complexity. We observe that refactoring scale is governed by the scope of the observable contract, the caller-visible interface fixing behavioral invariants while freeing private implementations for optimization. Building on this insight, we introduce FrontierRefactor, a benchmark of 60 qualified engineering tasks spanning four scales, from computational kernels to full repositories, across in-place optimization and cross-language rewriting. Evaluating frontier coding agents through a reference-free cascaded protocol across 1,080 execution runs reveals that correctness and speedup are largely decoupled: 41.2% of submissions satisfy functional equivalence yet fail to deliver verified performance improvements, with 19.3% suffering performance regressions and 21.9% trapped in measurement noise. Refactoring paradigms reveal orthogonal engineering bottlenecks: while cross-language rewrites gain speedups from native compilation after passing functional checks, agents struggle with contract fidelity, with pass rates dropping from 65.0% on modules to 32.6% at package scale; conversely, in-place optimization reliably preserves behavioral contracts at 86.5% but rarely achieves acceleration without physical performance profiling. Trajectory analysis shows that agents fall victim to uncalibrated micro-benchmarking illusions, an operational blind spot that multi-turn execution-timing feedback directly resolves, raising strict success from 16.0% to 40.0%.

★ A Cue's Hidden Price: Auditing Seller Influence on AI Shoppers

Felix Buecheler, Gerardo Berbeglia, Mingyu Guo — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Large language models are increasingly asked to shop on a user's behalf, which lets seller-controlled content such as badges, endorsements, and framing steer what they recommend. That poses the question at the core of generative engine optimization (GEO): not whether seller content shifts an agent's choice, but how much. The usual gauge of that influence, how often a cue wins the choice, has a blind spot: once a cue is strong enough to be picked almost every time, the choice rate can no longer show how much stronger one cue is than another. We introduce price-equivalent (PE) auditing, which removes this ceiling by pricing the cue in money. In a counterbalanced, position-controlled forced choice, we raise the cued product's price until it is chosen only half the time, and report that premium as the cue's price-equivalent: a formally defined, policy-conditioned parity threshold that makes very different cues comparable. Priced this way, cues that look equally strong at equal price differ by a wide margin, so choice-rate audits substantially under-resolve seller influence. We run the audit over 113,472 trials, seven models across families, and two product categories. To our knowledge, we also run the first price-equivalent audit of a decision-native model, Jev, which outputs a choice probability directly instead of generating text; strikingly, its profile is unlike any text model we tested, resisting most cues yet remaining susceptible to the strongest social-proof claims. We release the stimuli, logs, and code as a benchmark that reports each model's cue-price profile rather than a single score, giving developers and auditors a common, reproducible way to measure how content moves AI shopping agents.

★ When Should AI Chime In? Benchmarking Timely Contribution of Proactive Agents in Multi-Party Chat

Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Proactive agents can support group collaboration by surfacing useful information without waiting to be asked, but even relevant contributions can be disruptive if poorly timed. In this work, we isolate the timing question: given a fixed candidate contribution such as a clarification or relevant fact, when should an AI agent chime in? We introduce PITCH (Proactive Intervention Timing in CHat), to our knowledge the first benchmark with large-scale human judgments of proactive agent timing collected from participants in the same unscripted multi-party conversations they took part in. To collect high-quality judgments, we design a novel interface in which 2-4 people chat synchronously, author candidate AI contribution messages, place their own and others' contributions into the chat history, and then deliberate as a group about the best placement for each message. PITCH contains 384 candidate contributions and 1,091 judgments, each marking an ideal turn, an acceptable window, or that a contribution should not appear, from 89 participants engaged in 35 conversations. Evaluating 15 models across 28 configurations, we find that while models can place contributions well when given the entire conversation history, they systematically chime in too early when asked to decide live as the conversation unfolds: best-turn accuracy falls from 57.8% to 19.0%, with nearly half of placements occurring before every participant's acceptable window. In comparison, lay humans deciding live are more accurate (23.7% vs 19.0%), and fewer of their placements are premature (39.7% vs 46.4%). Model reasoning traces show that 93% of intervention justifications cite relevance, while almost none consider waiting. By providing a human-grounded benchmark, our work reveals gaps in agents' ability to integrate into group settings and points towards proactive agents that can reason explicitly about appropriate timing.

★ ClueFlow: Learning Long-Horizon Search through Attributed Research-State Transitions

Yicheng Wang, Wentao Tan, Yidan Huang, Yifei Liu, Qiong Cao — Tue, 20 Oct 2026 00:00:00 +0000

98% · Agents 98%, New model 92%, Efficiency 59%

Long-horizon search requires agents to acquire relevant evidence and integrate it into an evolving research state. Existing methods maintain research states to guide exploration or assign turn-level progress signals, yet grounding these signals in explicit, attributable state changes remains underexplored. We introduce ClueFlow, which learns long-horizon search from attributed research-state transitions. ClueFlow constructs Source-State-Final attribution chains that connect evidence-acquiring browser actions, Evidence and Candidate updates, and the final answer supported by those updates. During search, ClueFlow records typed state changes and potentially supporting browser actions, then applies gold-conditioned semantic validation and source tracing after execution. We use this attribution to guide both supervised fine-tuning and reinforcement learning. During SFT, attribution selects which decisions receive supervision with their original interaction contexts. During Turn-PPO, the same attribution augments outcome-based turn advantages with credit assigned along the Source-State-Final chains. Transition-level validation also recovers useful supervision from gold-consistent state changes in trajectories with incorrect final answers. We evaluate ClueFlow across two Qwen backbones and four search benchmarks spanning live-web and fixed-corpus settings, observing consistent gains from SFT and further improvements from PPO. With Qwen3.5-35B-A3B, ClueFlow achieves 60.0% accuracy on BrowseComp and 72.0% on BrowseComp-Plus. In controlled BrowseComp-Plus experiments, attribution-guided SFT matches all-turn exact-context SFT using about 30% of the training tokens, while attribution-guided PPO outperforms outcome-only PPO by 4.1 pp. These results demonstrate that state-transition attribution enables efficient supervision and effective credit assignment for long-horizon search agents.

★ FinHopper: A Progressive Multi-Hop Benchmark for Complex Financial Agents with Step-wise Evaluation

Yuxuan Wu, Miaopei Lin, Qiufang Ying, Lin Dan, Hao Zhao, Qirui Liu — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 97%, Hallucination 50%

Financial agents are expected to answer factual investment queries through open search, but financial multi-hop retrieval differs from general multi-hop QA in a crucial way: each intermediate answer is an exact typed constraint, such as an entity, date, event, or numerical indicator, that determines the next query. This strict answer-conditioned dependency makes small intermediate errors cascade into plausible but wrong search trajectories. We introduce FinHopper, a benchmark for complex financial agents with controlled reasoning depth from 1 to 6 hops, covering stocks, funds, and macroeconomic facts, with gold intermediate answers for process-level evaluation. To construct FinHopper at scale, we design a state-transition-based synthesis framework that composes verified atomic financial QA patterns into dependency-preserving chains, followed by targeted in-synthesis verification, model-ensemble checking, and public-search validation by financial experts. Experiments with 20 LLM-agent configurations show that public search tools are necessary but insufficient: performance drops sharply as hop depth increases, the best model remains far below human experts, and process-aware scoring reveals partial reasoning progress hidden by final-answer-only evaluation. The benchmark is available at https://github.com/justajustin/FinHopper/.

★ Honeyed Words, Hidden Daggers: Benchmarking and Training LLMs to See Through Mock Politeness

Caixin Kang, Sitong Gong, Tianyu Yan, Kaipeng Zhang, Yoichi Sato, Yifei Huang — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, New model 91%

The Chinese idiom "kou mi fu jian" ("honeyed words, hidden daggers") names speech that is polite in form but hostile in function. We present ZhenhuanBench, a 1,592-item benchmark built from expert commentary on Empresses in the Palace, covering speaker intent, emotion-language mismatch, and per-listener uptake. Text-only evaluation covers 34 LLMs. GPT-6 Astra leads the leaderboard at 73.48% overall, while its listener-reaction score is 58.37%. The annotations also supply supervised traces and string-verifiable training rewards. With 338 gold traces, an 8B model outperforms teacher distillation on identical items. Subsequent RL saturates, and chance correction reduces the estimated recoverable performance gap by a factor of 7.5. Our CPR-GRPO loop (Compress-Probe-Replenish) uses this estimate to weight gold supervision, reaching 67.0 on public multiple-choice tasks and 74.5 under the training-matched hidden protocol. Without target-task training, CPR outperforms Base, SFT, and GRPO across six Chinese and English pragmatic and sarcasm evaluations. On three external datasets, training with CPR yields higher mean Macro-F1 than SFT and GRPO.

★ From Code to Playable Systems: CEGame-Bench for Long-Horizon Game Creation Agents

Zixuan Dong, Yanghai Wang, Hanyan Bian, Jiaxi Tao, Junze Li, Jiadong Wang — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 70%

A game ships when a player can launch it, finish it, and return to a saved world, none of which a successful build guarantees, yet game-creation agents are mostly tested on one engine or on one-prompt browser games. We present CEGame-Bench, the first long-horizon benchmark for complete-game creation across five production engines: Defold, Godot, Roblox Studio, Unity, and Unreal Engine. Agents build each game from an empty project and a design document in runs averaging up to 295 steps and 65M tokens; the document doubles as a contract, and a hidden evaluator accepts a game only by real-input play. Under four scaffolds, agents earn most integration points but usually under half the gameplay points, and their games reach the ending in only 16% of runs, while references built by engineers with coding agents score near the maximum, which places this artifact-to-play gap in the agents. No two engines rank configurations alike (Kendall's tau 0.14-0.62), and the same game usually scores higher on Unity and Unreal Engine, which check more before play. We release Game Completion Boost, an agent skill for requirement-driven playtesting and repair that raises the execution-based score of three models by 14.1-24.3 points.

★ Reliable Before Profitable: QuantHarness for Governing Financial Strategy Development with LLM Agents

Jiasheng Mao, Jiacheng Liu, Hengwei Ye, Zheng Tian — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Hallucination 66%, Agents 61%

LLM agents can turn natural-language financial research requests into executable and backtested strategies, but a successful run may still implement the wrong rule, use information unavailable at decision time, or obscure how repairs and candidate selection changed the strategy. Because these failures arise throughout development, final code and performance metrics alone are insufficient to establish either fidelity to the request or the validity of the development process. We therefore address the reliability of financial strategy development at the trajectory level: the system must control how intermediate outputs become downstream state and preserve the evidence needed to justify the terminal outcome. We introduce QuantHarness, a two-layer agent harness in which a finance-specific Workflow structures specification, data grounding, construction, testing, and review, while Harness Core enforces stage-specific information, action, and commitment boundaries. We also introduce FinRelBench, a benchmark that evaluates strategy-development reliability through execution success, requirement satisfaction, and runtime evidence. QuantHarness achieves the highest Task Success at 90.72\%, compared with 74.23\%--84.54\% for the comparison systems. Controlled stress tests show that its boundaries contain targeted violations, while structured run records improve evidence retrieval and support decision replay under the evaluated protocol.

★ VEILED-SH: Vulnerability Expression through Indirect Language Evaluation Dataset (Self-Harm)

Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

AI systems may miss indirect, culturally situated distress or provide support mismatched to the application context. Existing benchmarks do not jointly address cultural variation, application context, and practitioner disagreement. We introduce VEILED-SH, a benchmark and construction methodology for evaluating single-turn responses to indirect expressions of suicide, self-harm, and distress. Its 10,003 prompts span counselor chat, forum posts, platform search, and triage, drawing on Singapore-grounded sources and public datasets. An expert-validated multi-label taxonomy distinguishes intent, communicative function, and response behavior. Six automated profiles, individually calibrated to three peer-support practitioners and three clinicians, retain separate reference policies for proxy-alignment scoring. Evaluating five systems on a coverage-enriched sample reveals variation in proxy alignment across profile-generated reference policies, even for the same system responses.

★ VibeTraining-Bench: Data-Trainer-Harness Co-Design for Iterative System Improvement

Ruotian Peng, Yuanyi Ren, Shuaiqi Liu, Xin Guo, Fang Zhu, Yangyi Huang — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 79%, New model 35%

Recursive self-improvement (RSI) requires agents to diagnose system bottlenecks and coordinate improvements across training and inference. We refer to the agents' capability to jointly optimize training data, the trainer, and the inference harness as cross-component co-design. We introduce VibeTraining-Bench, a benchmark for evaluating this capability across tasks spanning terminal operation, software engineering, multi-turn tool use, and mathematical reasoning. Within each task, research agents start from an identical initial agent system under a fixed resource budget. Guided by validation feedback, they iteratively refine data, trainers, and inference harnesses, then select and submit a final agent system for evaluation on the test set. To prevent reward hacking, we enforce three safeguards: test data isolation, separate research and grading containers, and network restrictions for research agents. Our evaluation results show that frontier research agents substantially improve target agent systems. For example, on SWE-bench Verified, GPT-6-Astra raises the target agent system's performance from 32.9% to 53.7%, comparable to the human-developed instruct model's 55.4%. However, there is still room to improve their ability to co-design and iteratively refine data, trainer, and harness to achieve cumulative performance gains. We hope that VibeTraining-Bench will serve as a unified platform for advancing RSI, deepening our understanding of cross-component optimization for agent systems. Code and data are available at https://github.com/D2I-ai/VibeTrainingBench.

★ RouterWorld: Scaling Conversational AI with Latency-Aware LLM Routing

Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Conversational AI systems built on Large Language Models (LLMs) require routing mechanisms that balance answer quality and response latency at scale. However, existing LLM routing benchmarks assume full observability of all candidate model outcomes, under-emphasize latency as a first-class metric, and fragment evaluation across heterogeneous protocols misaligned with real-world settings. We introduce RouterWorld, a comprehensive benchmark for LLM routing in conversational GenAI. RouterWorld unifies 371k single-turn responses and 76k multi-turn agentic trajectories, adopts natural user interactions with a standardized LLM-as-a-judge protocol, and decouples efficiency from transient pricing via hardware-grounded latency annotations. Crucially, RouterWorld formalizes routing under partial observability through two setups: Cold-Start, which restricts learning to sparse, static logs, and Voyager, which models continuous data streams where routing decisions shape future training distributions. Evaluating 13 routing baselines, we find that most routers over-emphasize accuracy while ignoring latency under partial observability. Code and benchmark data are publicly available.

★ StebBench: Diagnosing Real-World Procedural Understanding in AI Agents

Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Evaluating an agent's planning ability remains challenging for open-domain tasks, where verifying procedures through execution is often impractical. Existing evaluations typically assess only the final plan and do not distinguish procedural errors such as missing, incorrect, or misordered actions. We introduce STEPBENCH, a benchmark of 6,885 verified procedures crawled from wikiHow across 81 open-domain categories. From each procedure, we derive 5 diagnostic tasks: four with deterministic answers involving withheld, incorrect, or shuffled steps, and one requiring next-step generation. We additionally evaluate full-plan generation to examine how diagnostic performance correlates with end-to-end plan validity. Using the same procedures across tasks lets us directly compare different planning capabilities. For generation, STEPEVAL identifies critical failures that prevent goal completion. Across 14 configurations of 12 models, we find substantial differences across planning capabilities, with step ordering proving the most difficult: the strongest model correctly orders only 63.0% of procedures, dropping to 33.1% for those with nine or more steps. Incorrect-step detection also predicts judged plan validity across models, suggesting a lower-cost alternative for evaluating generation quality. On STEPBENCH-Val, comprising 214 procedures published in 2026, comprehension scores and rankings remain stable, suggesting limited influence from memorisation.

★ LG-Bench: A Graph-Structured Evaluation Benchmark for Life Sciences

Lu Sun, Xiangyi Zhang, Xiangyang Zhu, Kaiyuan Ji, Zijian Chen, Qiang Hu — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%

Traditional evaluation benchmarks reduce inherently interconnected scientific knowledge in life sciences into flat lists of questions, disregarding the underlying topological structure of the knowledge. We introduce, the first graph-structured benchmark for life sciences, featuring over 10,000 high-quality multiple-choice questions across medicine, biology, and chemistry. Our approach constructs a weighted evaluation graph using bidirectional matching and semantic similarity algorithms, where nodes represent questions and edge weights capture their semantic relationships. Leveraging this graph topology, we design two novel evaluation metrics. The Global Coherence Score (GCS) measures a model's consistency within semantically related neighborhoods, while Knowledge Balance Score (KBS) analyzes how model errors are distributed across the graph to reveal conceptual blind spots. LG-Bench facilitates fine-grained comparison of LLMs by surfacing differences in conceptual coherence and patterns of knowledge organization across models. Our framework shifts the evaluation paradigm from flat accuracy metrics to structure-aware analysis, offering a new lens for diagnosing and improving LLM performance in the life sciences domain.

★ RefCraft: Benchmarking Visual-Reference-Conditioned Agentic Generation through Web Development

Yuchi Wang, Rongyao Fang, Weihao Yu, Xilin Wei, Tiancheng Gu, Dingkang Yang — Tue, 20 Oct 2026 00:00:00 +0000

98% · Evaluation 98%, Agents 85%

Beyond basic tool use and coding skills, effectively incorporating visual references into agentic generation is equally essential for applying AI agents to real-world productivity tasks. We study this crucial capability through webpage development, a practical and readily evaluable setting in which existing benchmarks remain largely idealized: they typically focus on reconstructing visual targets or generating webpages from textual instructions alone, overlooking the more common workflow of developing from visual examples. Specifically, we introduce RefCraft, a benchmark comprising 300 realistic multimodal webpage-development tasks that combine natural-language instructions with screenshots from one to five reference webpages. Agents must selectively interpret and utilize these references while rendering, visually inspecting, and iteratively refining their outputs. We evaluate the resulting open-ended artifacts using independently verifiable keypoints and complementary programmatic and VLM judges. Across nine frontier (M)LLMs, the strongest system achieves an overall score of only 52.12 out of 100, revealing that visual-reference-conditioned agentic generation remains a substantial challenge for current models in practical webpage-development settings.

★ Do LMs Know What Variables Do? Interpreting Code Representations via Variable Roles

C\'elian Vasson, Daniil Ognev, Kentaro Inui, Benjamin Heinzerling — Tue, 20 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Variable roles describe recurring behaviors such as accumulation, traversal, and retaining the best candidate encountered. We use Sajaniemi's independently defined taxonomy to study which role distinctions can be recovered from large language models, how sparse autoencoders (SAEs) retain those distinctions, and whether associated directions support control of generation. We introduce static annotation rules for all eleven roles and a corpus of 126,926 natural-code files across five languages. Across five model-SAE systems, residual probes outperform individual SAE features and tested lexical baselines, with modest losses after neutral target renaming. Combining SAE features closes much of the single-feature gap, but probes on SAE reconstructions are less sensitive to role-changing edits reviewed by LLMs. High paired ranking accuracy coexists with frequent false positives at fixed thresholds. During generation, interventions often change role annotations at the cost of correctness; changes accompanied by passing tests show no clear aggregate advantage over random controls. The benchmark separates role-related recoverability, SAE fidelity, and correctness-preserving control through matched prediction, source edits, and generation interventions.

★ SteerSearch: Interactive Search Reasoning with Mid-Response Steering

Tue, 20 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Agents 96%, New model 93%

Search-augmented language models increasingly serve as long-horizon research agents, yet most open training methods assume a fixed user goal throughout search. We study mid-response steering, where users redirect a model during ongoing reasoning or retrieval. We identify answer-target anchoring, a failure mode in which fixed-query search agents retrieve evidence for the updated request but still answer an earlier question. We introduce SteerSearch, an RL framework that trains on interactive search trajectories with injected interventions to enable context-preserving target reassignment, and SteerSearch-Bench, a benchmark adapted from multi-hop QA datasets. We also investigate Contrastive Context Reward (CCR), an auxiliary signal for aligning post-intervention reasoning with the updated goal. Without CCR, SteerSearch improves average steering accuracy over Search-R1 from 11.1% to 31.3% on Qwen2.5-3B and from 36.6% to 44.6% on Qwen3-4B, while preserving standard single-query performance. Gains also extend to larger Qwen2.5 models and natural multi-turn conversations.

★ Not Every User Says It All: Training Tool-Using Agents Beyond Ideal Simulated Users

Han Bao, Zheyuan Zhang, Kaiwen Shi, Yanfang Ye — Tue, 20 Oct 2026 00:00:00 +0000

97% · Agents 97%, New model 92%, Evaluation 58%

Tool-using agents are trained and evaluated almost exclusively with strong simulated users who state every detail clearly, yet real users are often weaker. We construct weak users by masking sentences of a simulated user's replies, holding the task and the user's knowledge fixed, and cross-evaluate agents trained with strong and weak users on tau2-bench and ToolWOZ. Weak users lower the success of every policy, and training with them does not help: no policy trained with weak users, through fixed masks, mixtures or a curriculum, serves weak users better than the strong-trained policy beyond run-to-run noise, and prompting the agent to ask more does not help either. We propose StaGPO (State-Grouped Policy Optimization): the agent writes what the task needs, what is missing and how it is being filled, and GRPO uses this state to compare turns of different rollouts that reach the same situation, assigning step-level credit. Within the standard recipe of a small SFT cold start followed by RL, StaGPO makes weak-user training useful: on ToolWOZ, the Qwen3.5-2B policy trained with weak users reaches 76.8% with weak users, the highest of any Qwen3.5-2B policy we trained, 4.8 points above GRPO and 2.8 above the strongest baseline; it is also 5.4 points above GRPO with strong users, and generalises to an unseen mask rate; trained with strong users instead, it serves weak users at least as well as GRPO trained with weak users. On tau2-bench it scores highest with weak users for both models, by a small margin for Qwen3.5-4B.

★ The Judge Can Debias Itself: Awareness and Self-Debiasing for Reliable LLM-as-a-Judge

Youbing Yin, Jingyu Wu, Anoop Kumar, Scott Novotney, Daben Liu, Erin Babinsky — Tue, 20 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

LLM-as-a-Judge systems are increasingly employed for automated evaluation, yet remain susceptible to systematic biases that undermine their reliability. We demonstrate that the judge can effectively debias itself without additional training. We propose two lightweight strategies: bias-aware prompting, which explicitly alerts the judge to potential target biases prior to evaluation, and self-debiasing, in which the judge critiques the initial judgment for bias influence. To broaden the evaluation scope, we introduce three novel benchmarks--authority bias (fabricated citations), confidence bias (assertive phrasing), and formatting bias (structured presentation)--and evaluate across seven bias types in total. Across three frontier models, self-debiasing improves performance in 20 out of 21 model-bias combinations with gains up to 15 percentage points, while bias-aware prompting also yields consistent improvements. The recent fine-tuned external critic, however, does not generalize as reliably to the newly-introduced bias types (degradation observed in 5 out of 9 cases). Our findings demonstrate that prompting a judge to reason about bias, either preemptively or retrospectively, is a robust and generalizable approach for enhancing evaluation integrity.

★ CADS: Measuring the Incremental Contribution of LLM Proposers in Sequential Biological Screens

Guanxing Chen, Nihui Shao, Haohuai He, Zhenchao Tang, Pengwei Hu, Zhi-An Huang — Tue, 20 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Large language models are increasingly used to guide sequential biological screens, yet standard evaluations rely on raw hit counts that conflate a proposer's value with the discovery capacity of the underlying algorithmic harness. Adaptive sampling and corrupted experimental readouts also leave standard feedback comparisons without a false-discovery guarantee. We present Certified Agentic Discovery Screens (CADS), an error-controlled measurement framework that quantifies a proposer's incremental value against an identical LLM-free loop. By integrating capped likelihood-ratio bounds, e-processes, and complete-family e-BH, CADS controls the terminal false discovery rate under the stated null and corruption conditions despite arbitrary dependence, adaptive sampling, and optional stopping. In DepMap-backed simulations, CADS exposes two evaluation artifacts: the LLM-free harness alone clears the feedback gate, and raw origin counts overstate LLM contributions because LLM picks displace fallback discoveries. On the initial 20 seeds, supplying tested-candidate history to Qwen2.5-7B shifts its net contribution from +0.35, with a 95% interval covering zero, to +2.90, with the interval above zero. This initial net effect did not persist in a 50-seed matched-quota replication. These results support net gain over an LLM-free loop, measured between separately certified sets, as the appropriate measure of proposer contribution.

★ IndustrialCAD-1400: Benchmarking Detailed Industrial Assemblies from Text

Hebao Zhu, Kang An, Zhuoke Liao, Faqiang Qian, Ziliang Wang, Mingang Chen — Tue, 20 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Industrial assemblies require more than a recognizable shape: functional parts, interfaces, dimensions and service features must satisfy a detailed request together. We introduce IndustrialCAD-1400, a text-to-CAD benchmark for this industrial setting. Its 1,400 requests combine 400 equipment categories across 40 domains with 250 four-case design families. Family requests specify a median of 17 named component or configuration roles and six explicit numeric targets; controlled variants test whether dimensions change while shared structure is retained. Task construction draws on documented manufacturer information, owner specifications and engineering guidance. Models generate geometry programs from requests, and a common Blender environment executes them to produce scenes, optional measurements, and standardized views for evaluation. A request assessment of front, isometric, and top views measures visible component, arrangement, and interface fulfillment without requiring a single reference geometry. Five stratified panels preserve complete families while covering the benchmark exactly once, enabling balanced comparisons and separating industrial breadth from controlled numerical responsiveness. In conclusion, IndustrialCAD-1400 extends text-to-CAD from recognizable part geometry to industrial assembly synthesis under joint component, interface, spatial, and numerical specifications.

★ How Well Can LLMs Verbalize Mathematical Formulas? A Multi-Domain Evaluation

Peter Volbach, Tim Schlippe — Tue, 20 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Large Language Models (LLMs) have shown strong capabilities across natural language tasks. But their ability to generate accurate written verbalizations of mathematical formulas has not been systematically evaluated. Such verbalization is important for text-to-speech systems supporting audio-based learning and visually impaired users. We evaluated five state-of-the-art LLMs across six scientific domains and constructed Read-Aloud Math, a balanced multi-domain corpus with LaTeX formulas, manually validated simple and extended transcriptions, multiple variants, textual context, domain labels, and formula complexity levels. Using multi-reference BLEU, GPT-5.2 achieved the highest score for simple transcriptions (75.9% without context, 64.9% with context), although without context it did not significantly outperform Grok 4.1 Fast or Claude Sonnet 4.5. Claude Sonnet 4.5 achieved the highest score for extended transcriptions without context (36.5%), and Grok 4.1 Fast with context (48.8%). For extended transcriptions, Statistics and Mathematics scored highest across the two context settings, while Physics scored lowest. Formula length was significantly negatively associated with BLEU for simple transcriptions without context, but not with context, whereas BLEU increased significantly with formula length for extended transcriptions in both settings. Grouping symbols were most frequently involved in transcription differences, while variable names, capitalization, and Greek letters were associated with lower BLEU scores. Overall, LLM-based formula verbalization is promising but remains sensitive to transcription type, textual context, and formula characteristics.

★ Revealing what benchmark scores hide: Jointly measuring the breadth and reliability of LLM capabilities using accessibility profiles

Tue, 20 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Language model benchmarks assign each question a single outcome from one fixed wording, even though the same underlying request can be expressed in many equivalent ways. We argue that the quantity of interest should therefore not be correctness on one wording, but the accessibility of an inquiry, which we define as the probability that a model succeeds across meaning-preserving formulations. The distribution of accessibility across inquiries reveals a structure that accuracy and existing best-, worst-, or average-case prompt scores collapse: how broadly a model's capabilities extend across inquiries and how dependably those capabilities can be elicited. Two models can have the same average performance while one succeeds across more inquiries unreliably and the other succeeds across fewer inquiries but more consistently, leading to opposite preferences depending on whether the task allows multiple retries until one success or when repeated consistent success is required. Across six benchmarks and more than twenty models, this tradeoff is common: for about half of the model pairs with nearly equal mean accessibility, which model looks better depends on whether one success across five formulations is enough or all five must succeed, quantities we call Access@5 and Reliable@5, respectively. For some pairs the two measures diverge significantly, by up to 10 points in opposite directions. Accessibility also reveals that scaling, reasoning, and prompt optimization can redistribute capability across inquiries in ways aggregate accuracy obscures, sometimes broadening the set of inquiries a model can succeed on while making success on them less repeatable. Accessibility turns formulation sensitivity from a nuisance into a measurable property of capability, distinguishing what a model can sometimes succeed on from what users can actually depend on.

★ CorrectRTL: Boost RTL Generation Quality Through Automated Self-Correction with Instruction Decomposition

Jing Wang, Shang Liu, Hangan Zhou, Wenji Fang, Yuchao Wu, Yugao Zhu — Tue, 20 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Hallucination 39%

Large language models (LLMs) have shown strong promise in specification-to-RTL generation. Among the explorations, the self-correction paradigm has significantly advanced LLM-based RTL generation. However, two challenges limit further progress. Existing benchmarks are saturated by advanced LLMs, leaving little room to evaluate the agentic self-correction framework. The monolithic self-correction paradigm rewrites the entire design at once, where LLMs are susceptible to error confusion and corruption of originally correct code. To address the benchmark challenge, we contribute ScaleBench, a more challenging benchmark of 200 complex RTL designs. To address the challenge in methodology, we propose CorrectRTL, an agentic framework that realizes a new decompose-then-correct paradigm, in which the LLM agent explicitly decouples scaffold construction from detailed implementation. CorrectRTL contributes three core techniques: (1)~\emph{Structure-oriented decomposition}, (2)~\emph{Snippet-based RAG}, and (3)~\emph{SVA-based self-correction}. On ScaleBench, CorrectRTL improves Pass@1 from 20.5% to 49.0% (+28.5%). On VerilogEval, CorrectRTL achieves up to 96.1% Pass@1, demonstrating consistent improvements across both simple and complex designs.

★ One Skill Takes Over: Clashing Agent Skills and a One-File Fix

Parsa Razmara, Seyedarmin Azizi, Ardalan Aryashad — Tue, 20 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Agent Skills are instruction files that tell a coding agent how to do one job. Users install skills from many authors, so two skills for the same job can be active at the same time. We ask what a model does when both skill files are in its context: does it combine their compatible rules, or follow one skill and drop the other? We build SkillClash, 16 pairs of same-job skills in which code checks every rule and a reference output shows the best result that can be reached. On seven models from five families, from 4B open models to Qwen3-235B and GPT-4.1, a skill's rules are followed 93 to 99% of the time when it is loaded alone and 61 to 62% when a same-job skill is loaded next to it. Most outputs show takeover: every rule that only one skill asks for is followed, and every rule that only the other skill asks for is dropped. The worked examples inside the skills drive this. When each skill file instead ends with one example that follows both skills, takeover falls to 1 to 15% on the six models we tested. Telling the model which skill has priority does not bring the dropped skill back, and telling it to apply both skills brings the dropped skill back only on the two strong models and leaves the interacting rules broken. One merged skill file written by the strongest model we used, Claude Opus 5.5, lets all seven models follow 96 to 99% of the achievable rules. Most of its gain over a plain list of both skills' rules comes from its one worked example, and such an example helps only when it is correct. Code and data: https://github.com/prazmara/skillclash

★ Memories That Outlive Their Reasons: Dependency-Tethered Memory for Tool-Using LLM Agents

Yiqi Wang, Jinqian Ju, Jiaqi Zhang, Zhangkai Wu, Yiqun Duan, Zequn Sun — Tue, 20 Oct 2026 00:00:00 +0000

97% · Agents 97%, Evaluation 76%, Efficiency 73%

Agents now act on the Web, publishing packages, models and datasets through registry APIs, and remembering past verifications spares them the work. But a verdict rests on premises. When a signing key is revoked, everything built on it stops passing, yet the advisory names only the key, and the old verdicts of its dependents may never share a retrieval window with it. This non-locality of invalidation turns memory into a liability: in long simulated chats, an agent reading a memory of atomic facts publishes 62% of the objects whose failure lies upstream, each backed by a record that was true when written. Yet derivation is local even when invalidation is not. Tether records, at write time, the exact versions each verdict derives from, and at read time withdraws any verdict whose support has failed anywhere in the store; the agent reads the same fields as before. On Fallout, a generator of versioned worlds with hypothesis-blind simulated histories shared by every memory, Tether publishes 0.3% of upstream-failing objects across three domains when relations are stated with their verdicts, with 37% fewer verification calls than checking everything. A-Mem and Graphiti re-verify most candidates (0.83-0.86 checks per candidate), which limits their savings, yet publish 9% of upstream-failing objects; Mem0 relies on memory more and publishes 42%. Flushing every verdict on any failure is as safe as Tether but costs 32% more checks. The gain over fact memory holds with three open-weight backbones and against a whole-store window. These gains depend on capturing dependencies: when relations arrive in sessions apart from their verdicts, Tether's downstream unsafe publication rate rises to 9.4-22.1%, and a weak writer that misses edges loses safety in the same way.

★ SPARC: Spoken Profile Authorship Recognition Corpus

Tue, 20 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Speech anonymization has focused on the acoustic signal, leaving open how much speaker information the transcribed words themselves retain. In this work, we study pairwise speaker and demographic-attribute verification from transcribed speech. We introduce SPARC (Spoken Profile Authorship Recognition Corpus), a benchmark built from two podcast corpora comprising over 12,500 speakers, 1.5M quality-filtered utterances, and speaker-level inferred or source-provided labels for sex, age, and accent, and evaluate five attribution models and three LLMs on identity- and attribute-verification trials with episode- and metadata-based topic proxies. LLM adversaries distinguish same-speaker pairs from text alone. Attribute-label agreement is harder to verify, but LLMs perform sex- and accent-label verification above chance. Finally, releasing only short probe texts (50 words) degrades the strongest adversaries the most, offering a partial reduction in observed linkage.

★ Symbolic Constraint Agents: Mining Formal Tool-Call Constraints to Prevent Agent Errors

Jinu Lee, Shraddha Barke — Tue, 20 Oct 2026 00:00:00 +0000

97% · Agents 97%, Evaluation 55%

Symbolic constraints (programmatic, deterministic checks for tool calls) enable deterministic verification of LLM agents' tool calls, but existing approaches primarily enforce predefined safety policies rather than tool dependencies in complex tool environments. We explore whether symbolic constraints can be mined from environments and observations beyond predefined policies, and whether they can improve end-to-end task performance. We present Symbolic Constraint Agents (SCA), a framework that mines symbolic constraints from API documentation, refines them against execution traces, and checks the agent's tool calls with an SMT solver at runtime. A violation triggers a natural-language warning or, when the valid replacement is uniquely determined, an automatic repair. On AppWorld, refinement raises error-detection F1 from 37.0 to 58.3 on GPT-5.5 traces and from 39.3 to 61.9 on Qwen3.5-27B traces. At runtime, repair raises GPT-5.4-mini's task success by 5.4 points under a 40-call budget and 9.8 points when environment errors are hidden, while removing 59--87\% of failed calls across models, and transfers to $\tau^2$-bench with gains of up to 5.6 points. These findings show that automatically acquired symbolic constraints can improve tool-calling reliability and task completion, particularly when failures are costly or difficult to diagnose.

★ ToolOPD: Credit-Aware On-Policy Distillation for Tool-Use Agents

Chu Zhao, Lei Tang, Yanming Wang, Jianzhe Zhao, Guibing Guo, Zhengzong Chen — Tue, 20 Oct 2026 00:00:00 +0000

97% · Agents 97%, New model 94%, Evaluation 36%

On-policy distillation (OPD) enhances the capabilities of small language agents by providing dense, token-level supervision from a teacher model on student-generated trajectories. However, standard OPD implicitly assumes that credit is uniformly distributed across long-horizon trajectories, assigning the same strength of teacher supervision to different actions. Through credit-distribution analysis and outcome-reversal counterfactual experiments, we find that trajectory credit is concentrated on a small number of critical actions. Moreover, even when the trajectory is held fixed, reversing only the final outcome substantially changes the credit assigned to multiple interaction turns. These findings show that the uniform-credit assumption fails to identify critical decisions and ultimately degrades the model's reasoning accuracy. To address these challenges, we propose ToolOPD, an outcome-aware on-policy distillation framework designed for long-horizon tool use. ToolOPD first summarizes each completed trajectory and its outcome into a reusable hindsight skill. It then measures how this skill changes the teacher's probability of each student action token and converts the change into outcome-aware credit. The resulting credit preserves supervision that supports successful decisions or discourages failed ones, while suppressing irrelevant and conflicting signals. ToolOPD uses these credits to reweight the action-level distillation loss and jointly optimizes it with reinforcement learning. Extensive experiments across multiple long-horizon tool interaction benchmarks demonstrate that ToolOPD significantly outperforms existing reinforcement learning and self-distillation baselines.

★ Evidence Sufficiency and Model Comparisons in MexRegQA: An Item-Level Audit of Mexican-Spanish Regulatory QA

Tue, 20 Oct 2026 00:00:00 +0000

97% · Evaluation 97%, Hallucination 55%

Similarity-gated coverage does not establish that retrieved evidence contains the facts needed to answer a question. We audit Mexican-Spanish financial-regulatory QA and find that 35 of 135 covered items lack some or all load-bearing facts. On the post-hoc full-evidence slice (n=98), eleven prompt-matched models score 95-98/98 and none differs significantly from the anchor; this does not establish equivalence. Removing evidence-defective items changes the nominal significance status of 14 of 55 pairwise comparisons. On 56 zero-evidence items, accuracy spans 33.9%-89.3%. Removing evidence from the audited items under an abstention-permitting prompt reduces accuracy by 63-81 percentage points in all ten tested legs, with no recovered items. Building on sufficient-context research, we contribute MexRegQA v1.0: 192 items with provenance and evidence-status labels, frozen retrieval evidence for 191 evaluated items, predictions, judge outputs and offline checks. The audit was outcome-aware, the later model/panel sweep reused the same questions, and all reported scores and labels rely on machine evaluation without independent human validation. We propose releasing evidence-status labels alongside grounded QA benchmarks so that model comparisons can be examined separately from evidence availability.

★ A Calibrated Instrument for Detecting Disposition Drift in Language Models

Jerry Kaplan — Tue, 20 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Before language models are released, they are tested for harmful traits such as willingness to carry out dangerous requests, deception, power seeking, and sycophancy. The safe behavior these tests certify may degrade as models are modified, by topical fine-tuning, new training data, or changes to settings such as the depth of their reasoning. Most work reports such drift as single summary rates scored by another AI model, without the error analysis needed to interpret them. We adapt a calibrated measurement methodology recently developed for assessing how inference optimizations affect the quality of model outputs. Applying it to several misalignment dimensions in two open-weight models, we first establish the instrument's precision, showing that shifts of a few percentage points can be reliably detected. We then induce drift deliberately by fine-tuning on correct and on subtly harmful medical and financial advice; harmful advice induces a far greater shift than correct advice. Most strikingly, a model that reasons before answering can conceal the shift almost entirely in its answers, even though a direct reading of its internal state shows that its disposition has moved. With reasoning switched off, the same model shifts steadily as the share of harmful advice in its training grows. These results demonstrate the importance of testing safety with reasoning turned both on and off.

★ OpenSaC: A Python Library for Search as Code in Deep Research Agents

Qi Liu, Jiaxin Mao, Fengbin Zhu, Tat-Seng Chua, Ji-Rong Wen — Tue, 20 Oct 2026 00:00:00 +0000

97% · Agents 97%, Efficiency 84%, Evaluation 36%

Deep research agents investigate complex questions through iterative search, source examination, and evidence synthesis. A conventional workflow alternates reasoning with individual search and reading actions. Search as Code introduces a new paradigm: agents write programs that compose search, source reading, reranking, model calls, and local processing, returning selected observations for further reasoning. To support this workflow in existing coding agents, we present OpenSaC, an open-source Python library with a companion Search as Code skill. Configurable capability interfaces let agents combine these operations while retaining evidence for inspection and reuse. Compared with standard Codex using its built-in web search, the OpenSaC configuration uses less execution time and fewer model tokens on BrowseComp and WideSearch, with higher accuracy on BrowseComp and slightly lower row F1 on WideSearch. We aim to lower the barrier to adopting Search as Code and provide an open foundation for community development and transparent research on deep research agents.

★ GLOBE: Geometry-Led Online Behavioral Early Exit via Unit-Sphere Trajectories

Xinyue Peng, Zhenlin Liang, Meng Cai, Hua Wang — Tue, 20 Oct 2026 00:00:00 +0000

97% · Efficiency 97%, Evaluation 48%, Agents 30%

Chain-of-thought (CoT) reasoning can improve problem solving, but models often keep generating after the correct answer is already reachable; the extra steps raise serving costs and may replace a correct intermediate result with an incorrect one. We propose GLOBE (Geometry-Led Online Behavioral Early Exit), an online, training-free algorithm that terminates the model's CoT process early. GLOBE embeds each completed reasoning segment as a point on a unit sphere and reads the shape of the resulting semantic trajectory. GLOBE first applies a trial phase that generates short trial answers in a separate branch, stopping the main stream on an accepted trial after checking confidence against a threshold. If generation continues, a trajectory phase stops the stream once this path repeatedly reverses direction with little net progress. On five benchmarks including mathematical reasoning, tool calling, and agent memory, GLOBE cuts output tokens by 49.7% and active end-to-end wall time (E2E) by 51.0% across two models, and achieves a five-task average of 80.58%, comparable to Full CoT.

★ TokenGate: Adapting LLMs for Fast and Memory-Efficient Prefill with Learned Token Pruning

Liana Patel, Rishabh Ranjan, Ion Stoica, Matei Zaharia, Carlos Guestrin — Tue, 20 Oct 2026 00:00:00 +0000

97% · Efficiency 97%, New model 73%

Expert human readers process large inputs (e.g., code repositories and long documents) by skimming to closely read only what matters. On the other hand, language models process all input tokens through each model layer, resulting in significant processing costs for long-context inputs. We introduce TokenGate, a small, drop-in component that extends existing trained models to enable end-to-end learnable token pruning over long-context inputs for fast and memory-efficient prefill. To enable end-to-end learning despite the non-differentiable top-K operator for token pruning, we introduce token-gated attention, a mechanism inspired by Mixture-of-Experts that re-weights attention logits using token importance scores. Across diverse tasks, TokenGate achieves $1.4$-$11.3\times$ faster time-to-first token (TTFT) and $1.5$-$3.4\times$ lower peak memory at no accuracy degradation for prefill in transformers. Compared to the best-performing prior methods for token-pruning, TokenGate gives up to $6.4\times$ faster prefill and $2\times$ lower peak memory. TokenGate natively extends both standard transformers (e.g., Qwen3) and hybrid-attention models (e.g., Qwen3.5) with simple lightweight adaptation comparable to standard parameter-efficient fine-tuning. We make our code available at: https://github.com/tokengate-models/TokenGate.

★ ASTOR: Benchmarking and Improving Autoformulation for Soft-Constraint Optimization with a Formal Grammar

Amer Goel, Yeganeh Alimohammadi, Andrew Daw — Tue, 20 Oct 2026 00:00:00 +0000

97% · Evaluation 97%

Real-world decision problems combine two qualitatively different kinds of constraints: requirements that must be satisfied, called hard constraints, and preferences that can be violated, called soft constraints. Because these problems are often posed to decision-makers in natural language, large language models offer a natural tool for assisting with their solution. This has given rise to autoformulation, which focuses on using LLMs to translate natural-language descriptions of decision problems into executable mathematical optimization models. However, this research often neglects soft constraints like those proposed by consumers, managers, and planners, despite the fact that many practical problems require modeling them explicitly. We introduce ASTOR, an inference-time framework that maps a natural-language decision problem to a grammar-guided abstract syntax tree (AST) before deterministically compiling the resulting representation into executable optimization code. We also introduce SoftOR, a unified benchmark of natural language problems which focuses on soft constraint formulations. On SoftOR, ASTOR achieves the best aggregate accuracy among comparable single-request and multi-agent LLM methods while remaining competitive on standard autoformulation benchmarks that include only hard constraints. Code and data are available at https://anonymous.4open.science/r/astor-softor-iclr-E34C/README.md.

★ Agent World Model in the Wild: Computer-Use Agent Environment Synthesis

Zhaoyang Wang, Qianhui Wu, Haorui Wang, Simon Yu, Baolin Peng, Canwen Xu — Tue, 20 Oct 2026 00:00:00 +0000

97% · Agents 97%, Evaluation 85%, New model 75%

Agent learning is shifting from learning from human data toward learning through interaction with the world. However, scaling such learning directly in real-world computer systems is challenging and can introduce substantial security risks. We therefore propose Agent World Model in the Wild (WildAWM), a scalable pipeline for synthesizing controllable environments and tasks for training general computer-use agents (CUAs). Whereas existing approaches typically specialize in a single agent interaction interface, we construct executable environments that jointly support graphical user interfaces, model context protocol services, and command-line interfaces, allowing agents to flexibly combine visual interaction, tool use, and code execution. WildAWM is entirely powered by open-source models, demonstrating that high-quality computer-use environments can be generated without proprietary frontier models. Beyond environments, we address three recurring challenges in task synthesis: ensuring solvability, maintaining sufficient difficulty, and producing reliable verifiers, by introducing execution-based quality gates that combine heterogeneous agent rollouts, cross-judge consistency checks, and counterfactual probes of judgment robustness. The resulting pipeline synthesizes 341 long-horizon environment-task pairs that can require over 100 agent steps to solve and remain challenging even for frontier models, with GPT-6-Astra solving only 60% of them. We train two Qwen models of different sizes exclusively in these synthetic worlds. Both models achieve substantial gains across 8 out-of-distribution benchmarks spanning web, claw, and computer-use agents, demonstrating that learning in WildAWM's synthesized worlds generalizes broadly to unseen computer-use tasks. The code is publicly available at https://github.com/MSR-Orchard/WildAWM.

★ Same Simulation, Different Evidence: Where Retrieved Tool Output Costs LLM Agents Accuracy

Tue, 20 Oct 2026 00:00:00 +0000

97% · Agents 97%, Evaluation 84%, Hallucination 32%

LLM agents that compute the right answer can still answer wrongly. Agent harnesses increasingly keep long tool outputs out of the context window, saving them to files that the agent searches or reads on demand. We ask where this costs accuracy with a real scientific tool: PHREEQC, a widely used geochemistry simulator from the USGS, whose reports reach 846.8k characters and whose determinism lets every printed value be traced. Agents use it to answer 158 execution-verified multiple-choice items. When agents read the simulator's report through a section index rather than in context, GPT-5.1 and GPT-5.2 lose 11.3 and 7.9 points of accuracy, while Claude Opus 4.6 and Claude Sonnet 4.6 show no detectable cost. Re-running all 11,007 logged simulator calls shows that both GPT models compute the keyed value about as often as before; what changes is the evidence they answer from. Under retrieval they often skip the report, answer from their own higher-precision output files, and misround to a neighboring option, a failure we call evidence substitution. With the simulation held fixed, retrieval still costs GPT-5.2 9.4 points. For GPT-5.4, more frequent correct simulations mask an evidence-use loss of about ten points, leaving a net change indistinguishable from zero. Naming the authoritative output raises accuracy by 6.3-7.9 points over plain retrieval for the three GPT models; for GPT-5.1 and GPT-5.2, inlining outputs that fit works as well, while asking for correct rounding does not. Context-management choices in agent harnesses should be judged by where accuracy is lost, not only by final accuracy.

★ Factorization-Guided Training and Inference for Diffusion-Drafted Speculative Decoding

Jiakun Fan, Min Xiao, Qiang Wang, Neiwen Ling, Zhichao Liu, Zhihao Jia — Tue, 20 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, New model 80%, Evaluation 36%

Speculative decoding accelerates large language model inference through parallel verification of tokens proposed by a drafter. Diffusion drafters further parallelize proposal generation, but greater proposal capacity or higher mean accepted tokens (MAT) need not yield higher throughput. Because verification accepts only a consecutive prefix, early rejections can render later correct predictions ineffective, limiting MAT gains, while even higher MAT may not improve throughput if per-round latency increases. We introduce DScout, which uses two complementary factorizations to guide diffusion drafter training and inference for higher throughput. For training, DScout factorizes position-wise loss into inclusion, weighting, and prediction loss. This guides the design and combination of mixed-horizon training, intra-block unmasking, and acceptance-weighted loss to align supervision with prefix acceptance and improve MAT. For inference, DScout factorizes throughput as the ratio of MAT to per-round latency, guiding target-conditioning depth, drafter depth, and inference block size selection to identify a high-throughput configuration for each workload. Across six benchmarks, DScout achieves $3.84\times$ and $4.14\times$ mean speedups over autoregressive decoding on Qwen3-4B and Qwen3-8B, respectively, improving mean speedup over the state-of-the-art baseline by 9.1% and 7.1%. Across models and concurrency settings, our inference configuration policy provides up to 9.75% additional mean speedup over DScout's fixed configuration.

★ Analyzing Repetition and Interference Effects in Memory-Constrained User Simulators

Cara Su-Yi Leong, Nicholas Tomlin, Tal Linzen — Tue, 20 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

User simulators can help evaluate and train AI systems to interact with humans at scale. To do so, evaluation of user simulators must measure behavioral parity with people. We evaluate whether memory-constrained user simulators are sensitive to two factors known to affect human information retention: repetition of a fact and interference from competing information. We find that some simulators achieve human-like overall accuracy, and replicate the magnitude of repetition and interference effects, but none match people's accuracy on particular questions: while humans misremember facts, models forget them, leading to different error patterns. We release this data as a challenge benchmark for future work on user simulators.

★ FlashMap: Flow Map Speculative Decoding

Jeff Shen, Zhejian Zhou, Ryan Lee, Jonathan May, Shirley Ho — Tue, 20 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, New model 74%

Speculative decoding accelerates large language model (LLM) inference by using a lightweight drafter to propose a block of tokens that the target model verifies in one parallel pass. Autoregressive drafters generate this block one token at a time, whereas block diffusion drafters such as DFlash generate it in one pass, making drafting faster. However, these diffusion drafters predict each token independently, ignoring dependencies between tokens in the block. We introduce FlashMap, a continuous flow drafter that conditions a flow map language model on the target model's hidden states and generates the whole block in one forward pass while capturing inter-token dependencies. On Qwen3-8B, FlashMap is faster than a matched DFlash baseline on every benchmark, with 7% higher average speedup and up to 6.1x lossless acceleration.

★ Ask Less, Ask Better: Allocating User Attention in Proactive LLM Agents with Interruption Budgets

Tue, 20 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

LLM agents can ask clarifying questions to complete tasks, but answering each question demands users' time and effort. Existing evaluations typically measure this burden by counting questions, treating necessary and avoidable questions equally. We propose an evaluation protocol that distinguishes how often agents ask from how well they ask. An interruption budget limits the number of questions, while a user-attention cost charges extra for questions that are hard to read or not worth asking, as rated by a human-validated LLM judge. Beyond question counts, this cost captures perceived effort: on conversation pairs with the same number of questions, human annotators agreed with the cost on which conversation required more effort 75% of the time. Across eight LLMs and four customer-service domains, limiting agents to one to three questions reduces user-attention cost by 36-73% while keeping 86-94% of the task reward they earn with unlimited questions. At the same budget, reviewing each question against the memory of users' earlier answers cuts attention cost by another 13% with little reward loss and yields more valuable questions. Models also differ sharply in how well they ask: at the same user-attention cost, the most efficient models gain more than twice as much reward from their questions as the others. These results show that asking less and asking better are distinct goals, and that evaluations of proactive agents should measure both.

★ Muse Glimmer 30B: An On-Device Agentic Model

Tue, 20 Oct 2026 00:00:00 +0000

96% · New model 96%, Agents 92%, Efficiency 77%

We introduce Muse Glimmer, an open-source 30B parameter causal language model with perception encoder, purpose-built for autonomous agentic tasks on consumer hardware. Muse Glimmer integrates multi-step reasoning, reliable tool use and failure recovery into a single model that runs locally on consumer devices without requiring cloud infrastructure. The model accepts text and image inputs and produces text outputs, was trained on data from over a hundred languages, and is quantized to approximately 4-bit precision for efficient single-GPU inference. Muse Glimmer achieves state-of-the-art performance for its size class on agentic benchmarks, setting a new standard compared to the models of its size: Gemma4-31B and Qwen3.6-27B on tool use, multi-step planning, and agentic task completion. We release the model under the Apache 2.0 license to foster open-source development of capable, safe, local AI agents.

★ Means-Ends Analysis Helps LLMs Plan in Continuous-State Domains

Jiayi Li, Carl Kingsford — Tue, 20 Oct 2026 00:00:00 +0000

96% · Evaluation 96%

LLM planning research has traditionally focused on discrete-state domains, where actions form a finite, enumerable set. Existing LLM planning frameworks largely assume actions are enumerable, modeling planning as a search process over candidate actions. However, in continuous-state domains, where actions form a continuum and cannot be enumerated, this assumption breaks down. We find a remedy in Newell and Simon's Means-Ends Analysis (MEA), a general problem-solving technique rooted in heuristics from human psychology. We develop Medaka, an MEA-inspired structured prompting framework that encourages recursive goal decomposition, algebraic derivation, and per-step state tracking. To evaluate native LLM planning abilities in continuous-state domains, we develop Liquidworld, a systematic testbed on liquid mixing. On Liquidworld, Medaka can improve performance across model families, suggesting that classical means-ends analysis elicits LLM-native reasoning abilities that are helpful for continuous-state planning.

★ Auditing Tool Traces in Agentic LLMs: Source Localization and Counterfactual Dependence Testing

Junhao Liu, Tianxiao Li, Yongxian Wu, Ziheng Duan, Ren Li, Jing Zhang — Tue, 20 Oct 2026 00:00:00 +0000

96% · Agents 96%, Evaluation 93%

Agentic LLMs increasingly support scientific prediction by grounding decisions in external evidence. In biology, they promise accurate, interpretable predictions for important problems by integrating complex experimental and computational evidence. However, serialized tool results can contain hidden label shortcuts, such as dataset identifiers or template artifacts, that create apparent success without biological reasoning. To address this risk, we introduce a two-stage audit. Stage 1, Per-Tool TF-IDF Decomposition (PT2D), localizes label-predictive patterns to individual tool results while avoiding signal dilution in the concatenated tool trace. Stage 2, Token-Class Redaction, counterfactually tests whether the frozen LLM depends on the localized shortcut. We apply this audit to TCR-pMHC binding, where tool augmentation substantially outperforms a no-evidence baseline. At face value, this suggests that the LLM has learned to use scientific evidence. The audit overturns this interpretation: PT2D localizes the strongest signal to identifiers in one tool result, and redacting them reduces performance to chance. These findings show how hidden label shortcuts can masquerade as scientific reasoning or discovery, motivating source-level audits before benchmark gains are trusted.

★ LIRQuant: An Efficient Locally Isolated Rotation Quantization Method for Large Language Models

Desheng Xu, Hanglin Li, Tianchen Song, Zhiyong Zhao, Kun Zhan — Tue, 20 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, Evaluation 37%

Post-Training Quantization (PTQ) enables efficient deployment of Large Language Models (LLMs), yet activation outliers remain a key challenge for low-bit quantization. Existing global transformation methods overlook heterogeneous channel sensitivity to quantization noise, leading to suboptimal accuracy-efficiency trade-offs. To address this, we propose \textbf{Locally Isolated Rotation Quantization (LIRQuant)}, a sensitivity-aware framework that selectively applies transformations. LIRQuant identifies sensitive channels using a noise-weighted sensitivity metric, applying expressive learnable rotations only to these channels and lightweight fixed rotations to the remaining ones. This decoupled design reduces optimization complexity without sacrificing outlier smoothing capability. Extensive experiments across the LLaMA-2/3, Mistral, Qwen3, and Mixtral families demonstrate that LIRQuant achieves state-of-the-art accuracy under the challenging W4A4KV4 setting. Meanwhile, LIRQuant calibrates LLaMA-3-8B in only 25 minutes with 20.3 GB peak GPU memory, achieving the fastest calibration and the lowest memory footprint among recent learnable-rotation methods, and its fused inference kernels deliver up to $2.3\times$ prefill speedup and $3.5\times$ memory saving over FP16.

★ Heartbeat Thinking: Speculative Reasoning and Tool Execution While Listening in Realtime Voice Agents

Zhenyu Zhang, Kewei Fu, Yongqi Tong, Chenzhi Zheng, Wei Wu, Jiang-Ming Yang — Tue, 20 Oct 2026 00:00:00 +0000

96% · Agents 96%, Efficiency 81%, Evaluation 31%

Tool-using voice agents often leave users waiting after speech ends, even when actionable information arrives earlier in the request. Speculative approaches begin work during speech, but chunk-bounded or interrupt-on-update generation can truncate a tool call before its arguments are complete. This couples computation time to speech updates, limiting useful work completed in advance. We introduce Heartbeat Thinking, which handles updated input through new Thinker invocations while allowing ongoing invocations and dispatched tools to finish. Periodic invocations read the latest partial transcript and a shared record of completed reasoning and tool results. These results help resolve missing arguments and initiate dependent calls, allowing preparation to progress across speech updates. Across SpeechFCEval, FDBench v3, and AudioCRAG, Heartbeat improves aggregate speculative tool coverage or retrieval reuse over a SHANKS-style windowed baseline, with benchmark-dependent effects on final task quality. On FDBench v3, it increases tool coverage by 16 percentage points and reduces mean final-response latency by 36.4\%. The results show that heartbeat thinking paradigm preserving ongoing computation during speech can turn listening time into reusable preparation for faster and more capable voice-agent responses.

★ AfriGuard: Safety and Robust Refusal For African Languages

Tue, 20 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Efficiency 54%, New model 36%

Large language models (LLMs) designed to interact with users are typically trained to respond appropriately to harmful or toxic content. However, ensuring appropriate responses becomes more challenging in diverse linguistic and cultural contexts, where safety risks may be expressed or interpreted differently. Building safe AI systems therefore requires detecting safety risks, understanding their nature, and responding appropriately, while accounting for linguistic and cultural variations that add complexity to each stage. In this work, we introduce AfriGuard, a culturally grounded multilingual dataset for studying harmful intent, safety-category understanding, and refusal behavior across 10 African languages and English. Our findings reveal that guardrail models underperform both open- and closed-source LLMs on safety-intent and safety-category classification, while models show limited understanding of appropriate safety behavior, scoring below 60% F1. We further find that models often fail to respond appropriately to unsafe prompts not because they explicitly refuse, but because they produce linguistically inadequate responses that result in incidental non-assistance rather than proper refusals. Finally, using AfriGuard, we develop efficient guardrail models that achieve performance competitive with models up to 30x larger, introduce refusal detectors that improve average performance by 43.9% over existing detectors, and demonstrate that supervised fine-tuning can substantially improve overall safety performance.

★ Tandem: Serving Hybrid Linear-Attention GUI Agents on Mobile NPUs

Zhuomin Chen, Jingchao Ni, Xu Zheng, Yong Chen, Mo Sha, Wei Cheng — Tue, 20 Oct 2026 00:00:00 +0000

96% · Efficiency 96%, Evaluation 39%

Smartphone GUI agents complete tasks in mobile apps by repeatedly reading the screen and asking a model for the next action. Mobile neural processing units (NPUs) enable local execution, keeping screen contents on the phone, but model latency accumulates with every action. Hybrid linear attention models reduce memory requirements, yet reusing their recurrent state requires checkpoints at the appropriate input positions. Agent requests retain instructions but rewrite recent history, so checkpoint policies designed for chat continuation can miss reusable prefixes. Retaining rollback states for speculative decoding can also move prompt processing off the NPU's matrix unit, adding overhead that outweighs decoding savings on short action outputs. A loaded model can further hold the NPU awake between tasks. We present Tandem, a runtime for hybrid GUI agents on mobile NPUs. Tandem saves recurrent state where consecutive prompts diverge, allowing later requests to reuse previously processed context. It keeps most prefill on the matrix unit and produces rollback states through a short per-token tail. Its verification kernel reads each weight block once for all proposed tokens. Between request bursts, Tandem releases NPU power votes without unloading the model. We implement Tandem in llama.cpp and evaluate it on recorded agent interactions across two Snapdragon generations. On a Snapdragon 8 Elite, Tandem reduces per-step latency by 30% for text-based screen descriptions and 16% for screenshots relative to llama.cpp's fastest NPU configuration, reduces energy per text step by 33%, and lowers standby power with the model loaded from 0.76 to 0.22 W.

★ Parallel Language Models via Variational Decoding

Tue, 20 Oct 2026 00:00:00 +0000

96% · New model 96%, Efficiency 96%, Evaluation 36%

Autoregressive language models emit one token per forward pass. This causes high inference latency, in particular as models are scaled up for quality. In this paper, we introduce Parallel Language Models (PLMs), which generate mutually consistent tokens in one forward pass by routing dependencies through latent auxiliary variables. PLMs are trained using a variational bound that reduces to an autoregressive loss plus a parallel decoding loss. Empirically, PLMs are on a faster and better Pareto frontier than autoregressive and discrete diffusion baselines. Pretrained PLMs achieve near-autoregressive pretraining loss on model scales of 80M to 3B parameters, and speedup follows a power law in model size. When post-trained, our 1.3B model matches or exceeds an identically trained autoregressive model on math, coding, and instruction-following benchmarks, speeding up inference $2.3\text{-}5.7\times$. Together, we show that parallel generation can be trained from a single training objective at autoregressive quality. We will make our code available at https://github.com/mandt-lab/plm.

★ recAMPLIFY: Large-Scale Pretraining of Recursive Protein Language Models

Saurav Jha, Qianli Liao, Robert Vernon, Sarath Chandar, Quentin Fournier — Tue, 20 Oct 2026 00:00:00 +0000

96% · New model 96%, Efficiency 82%, Evaluation 31%

Protein language models (pLMs) are powerful tools for protein annotation and mutation-effect prediction, but progress has largely depended on scaling model size and data, making them expensive and prone to memorization. Test-time scaling offers a complementary path, but methods developed for autoregressive LLMs do not transfer directly: masked encoders lack a generation loop, and amino acids do not verbalize in a reasoning language. Instead, we adapt recursive pretraining to masked pLMs, where a weight-tied Transformer core refines residue-aligned hidden states over an adjustable number of cycles. We instantiate this recipe as recAMPLIFY, a pLM with 155M unique parameters trained on the same 2T tokens as AMPLIFY-350M. Despite 56% fewer parameters, it improves CASP contact recovery and three out of four supervised tasks, notably raising fold accuracy by 22.5%. Because optimal depth for recAMPLIFY varies across tasks and assays, we then introduce Gated Mixture Recurrence (GMR), a framework that decodes each protein from a probability-weighted mixture of hidden states visited during recursion. GMR employs a learned residual gate at scalar, low-rank, or per-dimension resolution to control how much of each cycle's proposed update is accepted into the state. This unifies halting and gating into a single adaptive computation objective while recovering PonderNet-style halting as a special case without gates. Layered on top of recAMPLIFY, GMR further reduces expected layer depth by up to 37%, although the accuracy trade-off varies by task. Mechanistic analyses show that residual gating controls recurrence stability beyond the trained depth and that halt depth tracks biological properties.

★ ProteinGeoBench: Frontier Structural-Biology Agents Are Fluent at Lookup but Lost in Geometry

Tue, 20 Oct 2026 00:00:00 +0000

96% · Evaluation 96%, Agents 92%

Frontier agents retrieve structural metadata well and compute geometry from coordinates less reliably. We ask how much of a benchmark score reflects that gap and how much reflects the evaluation configuration around the model. We study ProteinGeoBench, a suite of 131 protein-structure tasks, using 4,249 rollouts from four frontier models and 4,486 further trials across 26 models. Three of the four frontier models fall within 0.031 of one another (0.785 to 0.816) with no difference between them resolved, while the fourth scores 0.707 and all three gaps to it exclude zero. Three of the four do better on metadata lookup than on numeric calculation. The fourth is slightly better on numeric. Numeric tasks are 28 of the 131 but half of the twelve tasks that all four models fail. The ranking also depends on the evaluation budget: with one run per task, 23.9% of bootstrap replicates put a different model first. On a selected 15-task slice, removing network access lowers scores more than blocking eight structural-database domains, but the two contrasts were graded in different passes and cannot be cleanly decomposed. Swapping in each model's own command-line agent changes scores by -0.071 to +0.086, a comparison too small to rule out moderate harness effects. Across 26 models, numeric performance rises with non-numeric performance rather than staying flat. The experiments use different task subsets, so they cannot rank model, harness, access and task composition against one another. Evaluations should report results by task family, along with the number of runs, the harness and the access conditions.

★ OpenVLHarness: Real World Visual Agentic Tool-Use with Multimodal Memory

Yu Zhou, Cheng-Fu Yang, Rui Sun, Di Wu, Sibo Peng, Mingyang Zhang — Tue, 20 Oct 2026 00:00:00 +0000

96% · Agents 96%, Evaluation 79%, Efficiency 69%

Multimodal large language models (MLLMs) have made rapid progress on visual reasoning, yet many tasks require evidence that is difficult to obtain through internal reasoning alone: our controlled study on 22 benchmarks shows that simply increasing the inference budget of the base model yields only modest gains. Such tasks demand grounded perception, external knowledge, and numerical measurements obtained through heterogeneous tools, whose outputs must be made interpretable to the model and retained across reasoning steps. We introduce OpenVLHarness, a multimodal reasoning harness organized around a capability layer and persistent multimodal memory. The capability layer combines task-level tool interfaces with evidence rendering, providing consistent access to heterogeneous backends and converting their outputs into model-readable observations. Persistent multimodal memory retains the corresponding images, structured data, and web references as addressable artifacts throughout each query. A coding interface executes generated programs over stored data, enabling computations that combine results from multiple tools. Together, these mechanisms support an iterative process of acquiring, inspecting, and composing multimodal evidence. Evaluated on 22 benchmarks with eight open and proprietary backbones, OpenVLHarness improves average scores by 9.7-12.6 points over matched no-tool baselines, including +11.2 with Qwen3-VL-8B, +9.7 with Kimi K3, and +9.9 with GPT-6 Astra. With GPT-6 Luna at medium reasoning effort, OpenVLHarness outperforms the base model, Codex, and Visual Sketchpad at maximum effort by 7.7-8.5 points while incurring at most 37% of their model-token cost.

★ Web Agent Agentic Reinforcement Learning Decision Model Under Multi-Cost and Failure Risk

Qianli Ma, Limengxi Yue, Shuyang Xu, Yanpei Shi, Hongrui Liu — Tue, 20 Oct 2026 00:00:00 +0000

95% · Agents 95%, New model 53%, Evaluation 41%

Intelligent agent interactions in real-world web environments are commonly constrained by request budgets, time delays, anti-crawling restrictions, and operational failure risks. Strategies solely optimizing task success rates often exhibit unusable phenomena such as "high success but high cost" or "low risk but conservative failure." This paper proposes a constrained agentic reinforcement learning model for web agents, unifying page access, search requests, and external API calls into a unified long-term decision-making framework with associated costs. It simultaneously incorporates cost budget constraints and tail risk control into the optimization objective: constructing a multidimensional cost vector comprising cumulative request count, total latency, and failure penalties to achieve budget compliance via Lagrange dual updates, while employing a CVaR risk term to suppress excessive exploration of high-failure-probability paths, thereby achieving an adaptive balance among "completion rate, cost, and risk." Experiments were conducted across 30-70 site/page templates and 800-1,500 end-to-end web tasks, including information extraction, price comparison, form submission, and cross-page navigation. Interaction sequences spanned 20-120 steps with tool scales of 30-200. Performance was benchmarked against unconstrained RL, budget-constrained RL, and rule-based/scripted web agents, quantifying task completion rates, cost-per-success, failure rates, and policy stability. Results demonstrate that at equivalent completion rates, our method reduces C-PS by 22%-31% and lowers failure rates by 18%-26% under high failure penalties. Under fixed budgets, task completion rates increase by 10%-16%, highlighting the necessity and effectiveness of constraint modeling for practical web agent deployment.

★ Beyond Prediction: A Component-Level Study of Concept Recovery in In-Context Learning

Xufeng Liu, Fang Wan, Yi Liu — Tue, 20 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

In-context learning (ICL) enables language models to make accurate predictions from demonstrations, but prediction success alone does not reveal the extent to which the underlying concept has been recovered. Existing evaluations typically assess recovery at the level of the whole concept, making it difficult to identify which components are recovered and to separate model limitations from ambiguity in the demonstrations. We introduce a component-level framework for concept recovery that decomposes concepts into their defining components and uses controlled structural interventions to evaluate both component-level and whole-concept recovery under identifiable demonstrations. Across four instruction-tuned LLMs and three composition forms, we find that models often achieve high prediction accuracy despite poor whole-concept recovery, with systematic over-specification of relevant attributes and level constraints. Individual structural components are generally insufficient for whole-concept recovery, whereas jointly specifying complementary components, particularly relevant attributes and their level constraints, substantially improves recovery. However, this improvement is fragile: a single plausible but unnecessary level constraint can substantially disrupt recovery. Finally, joint supervision of recovery and implicit prediction substantially improves recovery while preserving prediction performance, with the learned capability transferring across composition forms. Overall, our component-level analysis provides a more diagnostic view of what models recover from demonstrations than the existing holistic view.

★ From SEC Filings to Financial Signals: Evaluating LLMs on Stock Market Reactions and Analyst Forecasts

Shengkun Wang, Taoran Ji, Linhan Wang, Sattar Mansi, Chang-Tien Lu, Yinlin Chen — Tue, 20 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

Form 10-Q MD&A sections combine long-form management narrative with accounting and forward-looking information. Benchmarks for financial filings commonly evaluate question answering or numerical reasoning, leaving unclear whether outputs from large language models (LLMs) align with market and analyst outcomes. We evaluate prompted, zero-shot outputs from 13 general-purpose and finance-specialized LLMs on 45,964 Form 10-Q filing events from 2001 to 2024, linking MD&A text to cross-sectional fundamental ranks, pre-filing analyst forecasts, and realized abnormal returns. The evaluation covers return ranking under five controlled information views and next-quarter EPS forecasting under direct and consensus-anchored interfaces. Relative to disclosure alone, adding cross-sectional fundamental ranks raises mean effective-date rank correlation (RankIC) by 0.0089 for effective-event-date returns and 0.0098 for following-day returns, whereas analyst-summary context provides no comparable average gain. Across the four disclosure-containing views, finance-specialized models show an effective-event-date group advantage but no statistically reliable following-day advantage, while general-purpose Qwen models produce the strongest individual return-ranking results. Under the full-information view, direct LLM forecasts average the 43.4th percentile of the analyst distribution and do not consistently beat individual analysts. Consensus anchoring raises the mean inside-range share from 41.1% to 59.5% but increases reported-EPS error.

★ An important Bottleneck in LLM Optimizers: Principle-Level Local Optima

Nian Ran, Tianhe Jia, Zhengyang Mu, Yue Wang, Sebastian Brulin, Markus Olhofer — Tue, 20 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

Feedback-driven LLM optimizers are increasingly used for evolutionary search, Bayesian optimization, heuristic design, and scientific discovery. Yet rapid initial gains often give way to prolonged plateaus despite substantial remaining evaluation budgets. We investigate \emph{principle-level search collapse}, where an optimizer's search becomes confined to a narrow set of design principles, even though it may continue to generate distinct candidates and achieve small improvements. We assemble a diagnostic benchmark spanning molecular optimization, network sentinel-node selection, and GPU-kernel optimization, and introduce a decision-tree protocol combining principle narrowing, prolonged performance plateaus, and gaps to attainable reference solutions. Across 161 assessable runs on the three problem domains, 78.3\% receive support for collapse, including 55.3\% with strong support. Candidate-level variation can nevertheless conceal restricted search at the principle level. We then test \emph{diversity injection}, a common response to search stagnation, such as higher t, multiple LLMs, and web search, and show that it provides limited and inconsistent benefits: among 65 molecular runs, still 73.8\% and 53.8\% receive support and strong support respectively. We finally test reference-principle interventions and find that they improve some runs, but their benefits vary across models and tasks: access to a better principle does not guarantee its effective use. Our benchmark and diagnostic protocol provide a basis for evaluating future optimizers on principle-level collapse. Our findings highlight the need for online collapse detection and better mechanisms for generating and maintaining alternative principles during optimization.

★ Screening LLM Judges Without Ground Truth: A Preliminary Evaluation of an AI Dental Hygiene Coach

Tue, 20 Oct 2026 00:00:00 +0000

95% · Evaluation 95%, Hallucination 74%

Evaluating a fine-tuned coaching model with expert graders is slow and expensive, and an LLM judge cannot be trusted without ground truth. We present a staged pipeline that screens candidate LLM judges and student models before expert review, using explicit decision rules at each stage: judge output audit, similarity with hosted judges, bias checks, and decision robustness. We demonstrate it on an AI dental hygiene coach. The case study uses synthetic, clinically unvalidated training and reference data to demonstrate the operation of the pipeline, rather than to establish the clinical performance of the resulting models. Two fine-tuned small language models (Granite-4.1-8B and Qwen3-8B), each trained at three data sizes, were scored on seven metrics by five judges (three local, two hosted). Two local judges, Qwen3.8-27B and Gemma-4-31B-it, were not distinguishable from the agreement between the hosted judges on six of seven metrics by rank correlation, and Qwen3.8-27B was the most similar to the hosted judges overall. No judge met our criterion for favoring its own model family or training-data source. The student comparison did not separate the two models on any metric under the pipeline's rule, so the choice of Granite-4.1-8B, which led in 29 of 35 judge-by-metric cells, is a lean that depends partly on truncated Qwen3-8B outputs. The pipeline also surfaced problems that raw scores hid, including the truncated outputs and hallucination flags from one hosted judge that no other judge corroborated. No dental hygiene expert has graded these responses, so the findings are preliminary and the pipeline's final stage, anchoring to expert judgement, is proposed but not performed.

★ TCMScience: An Autonomous Scientist for Traditional Chinese Medicine

Yanlan Kang, Ruiqi Liu, Xinpei Gu, Tzu-Yi Lei, Shuai Xu, Xukun Zhang — Tue, 20 Oct 2026 00:00:00 +0000

95% · Agents 95%, Hallucination 84%, Evaluation 57%

Traditional Chinese medicine (TCM) research connects classical texts, medicinal entities, molecular assays, computational predictions, and clinical studies. An autonomous research agent must preserve the inferential boundaries between these evidence types while coordinating analysis across heterogeneous resources. We introduce TCMScience, a scientific knowledge-discovery agent that separates language-model proposals from execution and release authority. Its architecture combines a trusted kernel, a scientific compiler, and claim-support contracts: plans are compiled into typed operations, execution is bounded by explicit permissions, and candidate findings are checked against source provenance, study design, and claim scope before release. The full capability registry contains 117 public-source connectors with 444 typed operations across 29 domains and 150 native tools across 12 domains; the TCM data hub catalogs 136 databases, 76 dataset specifications, and 84,294 formula-source records. These resources support model-independent workflows through browser and local execution interfaces. In a network-pharmacology analysis of Gegen Qinlian Tang, replacing the whole-proteome background with the assayed-protein background changed the number of significant pathways from 66 to zero; incorporating inactive screening records identified nine pathways, including drug-metabolism signals retained as mechanistic hypotheses. A governance ablation over 121 constructed erroneous outputs spanning 26 classes released 121 errors without governance, 57 with the domain layer, seven with the kernel, and zero with both layers. All 10 compliant controls were retained. These results establish complementary control of known failure modes and show how explicit evidence contracts change the conclusions an agent is permitted to report.

★ Toward Collective Model Intelligence: Characterizing and Coordinating Models

SinapisAI Team, Jian-Dong Liu, Qun Dai, Zhendong Du, Guannan Lai, Hao-Yu Liu — Tue, 20 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, Evaluation 87%

Independently developed AI models accumulate diverse and complementary capabilities, but a growing model collection does not automatically realize Collective AI. Addressing this gap requires understanding model capabilities with limited evaluation, organizing complementary models to collaborate on diverse tasks, and executing model calls efficiently under resource constraints. We present a systematic approach to pursuing Collective Model Intelligence through three connected components. First, Model Eval constructs evidence-based capability profiles by combining query-level performance estimation with cross-domain prediction, reducing the need for exhaustive evaluation. Second, Mio (Models in One) organizes complementary model candidates and supports task-conditioned, cost-aware selection and collaboration as task requirements evolve. Third, Mos (Model Serving OS) separates model selection from execution and provides endpoint scheduling and shared model deployment and distribution. Experiments show lower capability-estimation errors under limited evaluation. Across nine public benchmarks, Mio 1.0 Lite achieves the highest average score and lowest average cost among compared methods, with per-request costs 73.7-97.5% lower than those of the best-scoring alternatives. A prewarmed shared cache accelerates model-file retrieval by up to 23.3 times. This work lays a foundation for Collective Model Intelligence by characterizing and coordinating the diverse capabilities of an expanding model universe.

★ ReTRACE: Feedback-Guided LLM Agent in a Chemistry Harness for Retrosynthesis Planning

Jinxiang Xie, Yijian Zhang, Qingyuan Yi, Jianyi Wang, Hao Dong — Tue, 20 Oct 2026 00:00:00 +0000

95% · Agents 95%, Evaluation 87%, Hallucination 31%

Computer-aided retrosynthesis has advanced through single-step prediction, tree search and route scoring, but complex targets still require judgement over the whole route, where stereochemistry, functional-group compatibility, cost and greenness interact. Here we present \textbf{\mbox{ReTRACE}} (\textbf{RE}trosynthesis with \textbf{T}ool-orchestrated \textbf{R}easoning \textbf{A}nd \textbf{C}ontext-aware \textbf{E}vidence), an agent that places a large language model (LLM) in a chemistry harness that integrates molecular tools, complementary single-step prediction engines, a purchasable starting-material library, a route memory and a multi-criterion route evaluator. Routes are judged by a new metric that combines programmatic checks with adversarial LLM review, and its verdicts are compared with a synthetic chemist's. The agent learns continually from evaluator feedback during search, diagnosing weak routes and backtracking to revise them. With Claude Opus 4.8, at least one of the top 10 routes proposed by ReTRACE passed the audit for 87.4\%, 95.0\% and 86.7\% of targets on the three benchmarks. Two prospective wet-lab syntheses show that ReTRACE route proposals can be converted into executable procedures after expert revision.

★ Bellman Language Models: Solving the Recurring Decisions of LLM Agents

Adarsh Vatsa, William Eiers — Tue, 20 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, Agents 76%, Evaluation 58%

A general observation in agentic workflows, especially bounded ones, is that the decision space of the LLM, when we separate a decision from generation, starts approaching saturation: the agent meets the same states again and again, and after enough conversations it stops meeting new ones. LLMs generally don't treat these decisions as an optimization problem and instead re-derive them from their context/prompt on every turn. We show that treating decision making within reasonable bounds as an optimization problem, solving it once, and reusing the solved decisions can often make an agent faster, cheaper and more reliable. Our approach, which we call Discretize, Then Decide, starts from data such as the workflow's own logs or, when none exists, runs of the LLM agent itself. It reads each conversation into discrete states and estimates the probability of each outcome and the cost of each move, until no new states appear and the probabilities stop changing. It then solves the resulting Markov decision process. The solution runs as a controller around the model, sending moves whose content is fixed in the state, such as a tool call or a templated reply, and any state it hasn't seen is sent back to the LLM. We call an LLM run this way a Bellman language model (BLM), after Richard Bellman's equation for the best move in each state. On $\tau^2$-bench telecom, the controller with no model solves 79% of tasks. As a BLM, Qwen3.8-27B, an open model, goes from 29% to 96% with 93% fewer model calls. For Claude Opus 5.5, which solves 96% of tasks on its own, running it as a BLM cuts model calls and cost per task by about 98% and lets it solve 2 more tasks. On InventoryBench, the solved policy with no model beats every published method. Where the hard part is writing content rather than choosing moves, such as resolving SymPy issues, few decisions can be solved in advance, and a BLM helps little.

★ A Measurement Perspective on Benchmarking LLM-Generated Text Detectors using Item Response Theory

Yanbin Zhao, Zhipeng Chen, Guangsheng Bao, Guandong Zhang, Yan Li — Tue, 20 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

Benchmarking large language model (LLM)-generated text detectors commonly relies on aggregate detector-level metrics such as accuracy, F1 score, and AUC. Although useful for summarizing overall performance, these metrics provide limited insight into how benchmark composition shapes detector comparisons or which texts distinguish detectors with different response profiles. We propose a measurement-based benchmarking framework for jointly characterizing benchmark heterogeneity and detector behavior using multidimensional item response theory (IRT). By treating benchmark texts as measurement items and detectors as respondents, the framework estimates text-level discrimination and difficulty-location parameters alongside multidimensional detector abilities. We apply the framework to a benchmark of 32K human-written and machine-generated texts across four domains, evaluated by 14 detectors. The estimated IRT item-parameter space reveals distinct but overlapping human-machine distributions, with substantial variation across domains and text lengths. The human-machine label shows the strongest multivariate association with this space, while domain and text length further contribute to benchmark heterogeneity. Detector ability profiles reveal multidimensional and domain-dependent response patterns that are not fully reflected in aggregate performance rankings. As a practical application, clustering texts in the estimated IRT space yields a compact subset of 2,088 texts (6.5\% of the full 32K benchmark) that largely preserves the comparative detector-performance patterns observed on the full benchmark. Overall, the proposed framework provides a principled and interpretable methodology for benchmarking heterogeneous detection technologies while jointly characterizing the measurement properties of the texts used for their evaluation.

★ Conjugate Compilation: Self-Improving Agentic Co-Compilation for Speculative Decoding

Tue, 20 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, Evaluation 30%

As LLM adoption surges, scaling model serving requires unlocking stranded GPU capacity alongside deploying more GPUs. Speculative decoding brings two models into the serving process: a small draft model proposes tokens that a larger target model verifies. Yet compiling these models individually leaves their shared execution outside the compiler's optimization boundary. We introduce CONJUGATE COMPILATION, which reframes speculative decoding as a cross-model compilation problem and jointly optimizes the draft and target models within their shared execution rather than treating each in isolation. Our key insight is that cross-model compilation should be informed by the measured performance of the co-located shared execution it produces: an optimization that improves a kernel in isolation need not improve the models when they execute together. To expand cross-model opportunities, we devise operand layouts that interleave blocks from both models within common tiles, co-locating GEMM and non-GEMM work while preserving their mathematics. We realize CONJUGATE COMPILATION through a self-improving agentic search system in which an LLM edits cross-model optimization plans using the compiler and GPU execution as its environment. Each proposed plan is compiled and evaluated through its co-located shared execution, whose measured score, together with kernel-level verdicts and compiler feedback, informs subsequent proposals, while Monte Carlo tree search accumulates rewards to guide exploration. Across six single-GPU speculative-decoding configurations on NVIDIA B300 and RTX 5090 GPUs, CONJUGATE COMPILATION achieves a geometric-mean per-round speedup of 1.37x over conventional execution. Interleaved kernels raise this speedup to 1.45x, reaching 1.59x in the best case. CONJUGATE COMPILATION also improves GPU utilization on average by 1.46x on RTX 5090 and 1.50x on B300.

★ UpLoop: Upcycling Large Language Models with Gated Loop

Tieyuan Chen, Huabin Liu, Yingtao Zhang, Shijie Li, Tengyu Pan, Weiyao Lin — Tue, 20 Oct 2026 00:00:00 +0000

95% · New model 95%, Efficiency 36%, Evaluation 31%

Looped large language models trade parameter storage for recurrent computation, but most are trained from scratch. In practice, pretraining with looped depth is expensive, making post-training upcycling of a pretrained non-looped LLM an attractive alternative. We introduce \textbf{UpLoop}, a post-training recipe based on \emph{layer-wise} recurrence. Instead of looping a whole model or a middle block, UpLoop loops each selected layer locally, preserving the pretrained depth hierarchy and avoiding feature-space mismatch. To make loop execution useful, we add hidden-wise gates on the attention and MLP residual branches, so later loops act as progressive refinements rather than redundant rewrites. Empirically, UpLoop breaks the commonly observed saturation at two loops ($R=2$) in prior looped models. Starting from Qwen3-4B Base and applying SFT, UpLoop improves the average accuracy across 20 benchmarks by 6.17\% at $R=3$. We use $R=3$ as the main setting for the best accuracy--cost trade-off, while both native and extrapolated $R=4$ yield further gains.

★ Fine-Grained Units, Flat Decisions: Hierarchical Micro-Expert Pruning for Mixture-of-Experts Models

Zukang Xu, Xinpei Gao, Xing Hu, Zhe Jiang, Dawei Yang — Tue, 20 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, Evaluation 43%

Micro-expert pruning reduces the capacity of mixture-of-experts (MoE) models by removing coupled channels within routed experts. However, finer pruning granularity alone does not guarantee the preservation of downstream capabilities, as similar language-modeling perplexity can accompany substantially different task performance. We revisit micro-expert mask construction as two distinct decisions: allocating retained capacity across experts and selecting channels within that capacity. We propose UC-RAMP, a two-stage post-training pruning framework that explicitly separates these decisions. CAPE uses cross-domain calibration evidence to redistribute expert widths under an exact layer-wise budget, while RAMP refines channel composition at fixed widths through one-out, one-in exchanges guided by a channel-interaction surrogate. Across three MoE models at 25% and 50% routed-expert channel pruning, UC-RAMP achieves the highest ten-task mean accuracy among evaluated pruning methods in five of six settings. At 50% pruning, it outperforms the strongest evaluated baseline by 8.19, 3.96, and 6.14 percentage points across the three models, respectively. In a single-A100 deployment at 25% pruning, UC-RAMP reduces measured time to first token by 19.7--23.9% and time per output token by 20.2--22.3% relative to the unpruned model. The code is available at https://github.com/xzktx003/UC_RAMP.

★ TimelineAgent: An LLM-based Framework for Query-Centric Event Timeline Extraction Across Multiple Documents

Qingkai Min, Yunxin Bao, Yifeng Lu, Zhenyu Feng, Yulong Chen, Andreas Vlachos — Tue, 20 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

Query-centric cross-document event timeline extraction aims to construct event timelines from multiple documents under diverse query focuses. The task is very challenging due to the heterogeneous event modeling and reasoning requirements arising from both the intrinsic complexity of cross-document timeline construction and the diversity of query focuses. In this paper, we propose TimelineAgent, an LLM-based multi-agent framework featuring a multi-stage reasoning workflow, where heterogeneous event features are organized into reusable skills and orchestrated through a goal-oriented protocol. To facilitate evaluation, we construct a benchmark covering diverse query focuses. Experimental results show that TimelineAgent consistently outperforms existing methods, offering a practical framework for complex real-world timeline construction.

★ The KV-Surgeon: Compensation-Aware Offline KV-Cache Compaction

Moritz Wagner, Christophe Roux, Max Zimmer, Sebastian Pokutta — Tue, 20 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, Evaluation 38%

Offline compaction of the key--value cache (KV-cache) reduces the cost of repeatedly querying a fixed context by constructing a smaller cache once and reusing it subsequently. The central challenge of compaction is deciding which cache entries to delete and which to retain. Existing methods score each candidate for deletion on its own, with the rest of the cache being held fixed. However, softmax renormalization of the attention operation couples these decisions in a nonlinear manner: removing one entry changes the attention weights of every remaining one. In this work, we introduce KV-Surgeon, which uses calibration data to take interactions into account in a tractable way. We derive a quadratic surrogate that admits closed-form expressions for jointly making the optimal greedy removal decision as well as compensation within the remaining entries. On QuALITY, LongHealth, and Long Code Arena, and at every evaluated compaction ratio, KV-Surgeon substantially advances the cost--accuracy Pareto frontier of dense and mixture-of-experts models from 4B to 30B parameters. At $50\times$ compaction, KV-Surgeon improves question-answering accuracy by up to four points over attention matching (Zweiger et al., 2026), the state of the art, at comparable cost. Lastly, we introduce an even faster variant, which also exceeds the state of the art while requiring only two fifths of its cost.

★ ReAnchor: Reusing Rejected Verifier States for Diffusion Drafting

Sanghyun Lee, Wonjin Nam, Seungho Jang, Jaeho Lee, Jiho Park, Jaewoon Lee — Tue, 20 Oct 2026 00:00:00 +0000

95% · Efficiency 95%

Block diffusion drafters accelerate speculative decoding by generating multiple draft tokens in a single forward pass. However, each pass produces only a fixed-length block from a single target-generated anchor, limiting how far the draft can extend. Meanwhile, target verification computes hidden states even at positions beyond the first rejection, but these states are typically discarded rather than reused. We propose ReAnchor, a training-free method that reanchors a pretrained drafter at shifted positions using discarded verifier states to generate parallel draft blocks. These blocks improve predictions within the original block and extend coverage beyond its endpoint. At sampling temperature of T=0, ReAnchor achieves its largest average gains on Gemma 4-12B, improving acceptance length by 19.4% and decoding speed by 17.9% across eight benchmarks. Averaged equally over four target models and eight benchmarks, it improves acceptance length by 15.2% and decoding speed by 13.4% with pretrained DFlash drafters.

★ Breaking Diffusion: Parallel Reasoning with Draft Refinement

Jaesang Yu, Seoyeon Byun, Yun Wang, Yan Lu, Juho Kim — Tue, 20 Oct 2026 00:00:00 +0000

95% · New model 95%, Evaluation 64%, Efficiency 48%

Drafting an answer allows a reasoning model to reconsider each decision in the context of the others. Yet repeated diffusion denoising does not ensure that individually plausible token updates form a better answer together. We introduce Draft Refinement Language Models (DRLMs), which use a single denoiser to propose parallel revisions and evaluate a candidate draft. We train the denoiser to reconstruct every token from uniformly corrupted drafts, teaching both preservation and repair. Its predictions specify revisions directly, without a diffusion reverse transition. Evaluation requires a different use of these predictions: confidence in an existing token reflects both contextual support and evidence from observing the token itself. Using the known corruption process, we correct for this self-token evidence to estimate how well each token is supported by the prompt and the remaining draft. Aggregating these estimates yields a cavity energy computed in one forward pass, without an additional scoring model. Each round proposes a revision, reads the completed candidate in its revised context, and retains it only if its energy is lower. Across Countdown, Sudoku-Extreme, and Maze-Hard, our method outperforms autoregressive and discrete diffusion baselines. With 6M parameters, DRLM reaches 44.4% on Countdown-5 and 93.9% on Sudoku-Extreme. The results establish complete-draft parallel refinement as an effective alternative to reverse diffusion for complex reasoning.

★ WhatIf-Bench: Benchmarking Intervention-Aware Time Series Forecasting

Tue, 20 Oct 2026 00:00:00 +0000

95% · Evaluation 95%

Recent time series foundation models (TSFMs) can consume covariates in-context, extending zero-shot forecasting to problems driven by external factors. This enables scenario analysis: forecasting how a target responds to events that set covariates to chosen future values. Whether they can do this reliably is an open question. We introduce WhatIf-Bench, a benchmark measuring how well forecasters use covariates to forecast a target's response to future events. WhatIf-Bench is semi-synthetic: time series come from mechanistic simulators of data centers, power grids, road traffic, and water networks, with controlled covariate-target relationships grounded in each system. Each task presents a history of events intervening on covariates, then asks to forecast a future event that either replays seen values (Replay) or sets them to unseen ones (Novel). Varying the number of covariates, historical events, and event length yields 37,665 tasks. Evaluating TSFMs, frontier LLMs, and simple regression baselines, we find the LLMs reach the highest skill (improvement over a naive baseline), ahead of the best TSFMs. Every model forecasts well on Replay futures (48 to 85% skill) but drops to -7 to 50% on Novel futures, with the regression baselines falling the most. Skill on Novel futures also falls as events set more covariates jointly. More historical events raise skill on Novel futures across the board, but large differences remain. WhatIf-Bench reveals generalization to Novel futures as the core challenge, and offers a testbed toward arbitrary scenario analysis.

★ MemT-1.0: Towards Memory-Native Evolution

Zhiyu Li, Zeyu Zhang, Linyun Liu, Yiwei Bu, Chenyang Xi, Xichong Zhang — Tue, 20 Oct 2026 00:00:00 +0000

95% · New model 95%, Efficiency 92%, Evaluation 49%

Memory-native evolution requires making memory an intrinsic capability of foundation models, enabling them to continuously accumulate and utilize information. To this end, we introduce MemT-1.0, a new architecture of memory foundation models that internalize memory as part of model computation. Building on Metis, MemT-1.0 employs Gated DeltaNet-2 with a stable token-level memory update mechanism within each turn. We further introduce Mixture-of-Storage to scale native memory capacity. Following a deliberate backbone preparation, we develop a two-stage training pipeline consisting of continued pre-training and post-training, supported by 34B tokens of training data and efficient training and inference infrastructure. We release a model family comprising MemT-1.0-Nano and MemT-1.0-Flash, and conduct a comprehensive evaluation of memory and general capabilities. The results show that they outperform the parametric memory baselines on average and remain competitive with retrieval-augmented methods, while largely preserving general capabilities. At a history length of 128K tokens, MemT-1.0-Nano achieves approximately $1.86\times$ end-to-end inference speedup and a $66\times$ reduction in persistent storage relative to full context KV cache. By internalizing retrieved procedural experience, MemT-1.0 improves task success on agentic tasks, providing a practical step toward memory-native evolution.

★ LatComm: Reusable Context Compression for Latent Communication between Large Language Models

Huawei Lin, Jinting Liu, Tan Yu, Weijie Zhao — Tue, 20 Oct 2026 00:00:00 +0000

95% · Efficiency 95%, Evaluation 41%, New model 31%

Systems that combine several large language models (LLMs) often use one context for many questions. When the context is passed between models as text, the receiving model processes all of its tokens again for every question. Avoiding this cost requires a context representation that is compact, reusable across questions, and readable by a different LLM. Existing compression and latent communication methods each provide only some of these properties. We introduce LatComm, a Latent Communication interface that provides all three by separating encoding from reading: the context is encoded once without the question, and the receiver selects what to read for each question. A frozen context encoder LLM compresses the context into a fixed-size packet and sends it to a different frozen receiver LLM, which reads the packet through cross-attention, with gates computed from its prompt. On 12 multiple-choice question-answering datasets, LatComm improves average accuracy over single-vector baselines by up to 10.2 points. On QuALITY, it reduces the receiver's input by 57x and answers the questions of an article 3x faster.

SAGE-CP: When Marginal Coverage Hides Route-Conditional Failures in Adaptive VLM Ensembles

Mushir Akhtar, M. Tanveer — Tue, 20 Oct 2026 00:00:00 +0000

94% · Efficiency 94%, Evaluation 41%

Adaptive VLM ensembles save computation by stopping early on easy images and escalating harder ones. Although a single conformal threshold can make the final predictor marginally valid, it can still fail severely along individual computational routes. We characterize this failure precisely: if a route has prevalence pi, marginal 1-alpha coverage allows its conditional coverage to fall to max{0, 1-alpha/pi}. To address this mismatch, we learn the router on development data, freeze all routing choices, and calibrate the returned score separately for early exits and escalated inputs. We call this pipeline SAGE-CP. Its calibration step applies Mondrian split conformal prediction to the operational groups created by the router. Under exchangeability, the resulting prediction sets achieve finite-sample coverage within each route. Across 10 benchmarks, 4 frozen VLMs, and 5 aligned splits, both global calibration and SAGE-CP achieve 0.900 marginal coverage at a 0.90 target, but their route behavior differs substantially. SAGE-CP raises mean worst-route coverage from 0.623 to 0.887 and reduces the mean route-coverage gap from 0.324 to 0.025, using the same routed scores and 1.177 source calls per image. The stronger guarantee increases the mean prediction-set size from 1.893 to 2.212; nevertheless, the sets remain smaller than those of the full ensemble (2.486) while requiring 70.6 percent fewer source calls. The route-level reliability persists across 6 fixed query fractions, while finer exit-depth groups demonstrate the need for additional calibration data on rare routes. Across 5 datasets, SAGE-CP also reduces measured batch-one latency by 58.8 to 63.4 percent and gross GPU energy by 55.7 to 57.5 percent relative to full inference. Thus, routing provides computational savings, while route-stratified calibration attaches coverage guarantees to the operational paths created by the router.

Doing Less with More: A First-Principles Exploration of the Suitability of Agentic Computing over Alternative Architectural Choices

Ritvik Garimella, Biplav Srivastava, Amit Sheth — Tue, 20 Oct 2026 00:00:00 +0000

94% · Evaluation 94%, Agents 59%

There is growing interest in automating business activities with Agentic Artificial Intelligence (AI) due to latter's seeming ease of use. Never has it been easier, or costlier, to do less with more. However, little is known about when agents are preferable to established alternatives such as local computation, Representational State Transfer (REST), the Simple Object Access Protocol (SOAP), and the Model Context Protocol (MCP), particularly when development speed, performance, and operational cost are considered. We investigate this question using a controlled mathematical task that compares seven methods on a benchmark of 1,000 arithmetic expressions where semantics of operator precedence has to be preserved. We ran this setup across a native Function Calling baseline (Python library), REST, SOAP and MCP microservice architectures, direct responses from large language models (LLMs) using a Groq cloud API and five locally deployed Ollama models, and agentic systems in which five local LLMs invoke MCP tools. The deterministic methods achieve 100% accuracy with negligible latency. Direct LLM evaluation achieves at most 92.9% accuracy at over 22,000x the latency of a native function call. With MCP tools, the best agent reaches 99.9% accuracy, but incurs nearly 1,000,000x the baseline latency. These results demonstrate that LLMs approaches are poorly suited for exact arithmetic evaluation and motivate an informed architectural decision framework (ADF) for choosing an architecture based on accuracy and throughput requirements. We recognize that an actual interaction with an agent may span both reasoning and linguistic tasks. Hence our recommendation is to explore multiple architectural choices (based on task type) rather than one-size-fits-all solution (LLMs). Code is available at - https://github.com/Ritvik-G/adf/

LatLM: Line-Parallel Language Models with Deferred Commitment

Shaorong Zhang, Shiyuan Li, Longxuan Yu, Greg Ver Steeg — Tue, 20 Oct 2026 00:00:00 +0000

94% · Efficiency 94%, New model 93%, Evaluation 47%

Line-parallel language models advance multiple lines concurrently for substantial speedups. However, frameworks like the Line-Coupled Language Model (LCLM) enforce a fixed commitment schedule across lines. Because dependencies vary within text, a fixed schedule cannot hold back a line whose prerequisites remain uncomputed, forcing premature commitments and cascading reasoning errors. To address this limitation, we introduce the Lattice Language Model (LatLM), making commitment schedules an adaptive decoding-time choice. Text occupies a 2D lattice where eligible frontier tokens are scored each round; a policy commits confident candidates while deferring unready ones for extended context. One forward pass supervises entire sampled trajectories via 2D causal masking, while an append-only cache retains exact states. At 482M parameters, we establish two findings. (1) LatLM matches a data- and capacity-matched autoregressive model on held-out text when decoded sequentially, while its adaptive parallel decoder achieves near-autoregressive accuracy on independent-line tasks using 25--41% as many decoding rounds. (2) When cross-line dependencies cause LCLM's fixed schedule to fail, deferred commitment substantially improves accuracy while retaining parallelism where dependencies permit. On mixed dependency chains, LatLM achieves 96.5% accuracy versus LCLM's 78.5%, using fewer than half the decoding rounds of the autoregressive baseline.

Prosodic Structure as a Critical Component in Phonemic Language Modelling at BabyLM Scale

Bill K. S. Ku, Janet B. Pierrehumbert — Tue, 20 Oct 2026 00:00:00 +0000

94% · Evaluation 94%, New model 47%

Human children learn language from continuous, prosodically structured speech long before they can read, yet almost all language models train on orthographic text. Building on previous phonemic BabyLM training (Goriely et al., 2024; arXiv:2410.22906), we ask whether supplying an estimate of the prosody infants exploit improves phonemic language modelling at BabyLM scale. We compare an orthographic baseline to three conditions forming a ladder of phonological richness. All conditions are trained as BERT MLMs at 10M and 100M words. Addressing the ecological validity of existing metrics in the small-data regime, we build a lexical decision task and two semantic organisation tasks. Our new lexical decision task offers improved insights into the distinction between wordlikeness, token familiarity, and lexicality as factors. Triphone phonotactic control and pseudo-log-likelihood scoring shows that these scores reflect phonotactic familiarity rather than wordhood: every model sits near or below chance on an adversarial inverted condition, revealing prior results of this kind as partly artefactual. On our new semantic tasks, a prosodically annotated phoneme stream gives the strongest taxonomic organisation at both the 10M and 100M word scales. Phonemic models with boundary structure are competitive with the orthographic baseline. These findings suggest that prosodic structure is a key component of phonemic language modelling in the small-data regime most analogous to infant learning.

Training-Free Prompt Selection for Efficient Reinforcement Learning With Partial Rollouts

Lei Jiang, Honghui Bao, Yue Zhou, Zhengyao Gu, Langzhou He, Natalie Parde — Tue, 20 Oct 2026 00:00:00 +0000

94% · Efficiency 94%, Evaluation 58%, New model 38%

Reinforcement learning with verifiable rewards (RLVR) has emerged as a prominent paradigm for training reasoning models, but its computational cost makes efficient data selection increasingly important. Since training examples contribute unequally to learning, selecting a small, informative subset can offer a more efficient alternative to using the entire data pool. The training-free selection method estimates prompt value with hidden states from complete greedy rollouts. However, in our datasets, $26$--$32\%$ of these rollouts fail to terminate within the generation budget. We introduce LIFT, a training-free data selector that identifies impactful prompts using only partial reasoning rollouts. At its core, LIFT uses Partial Rollout Context Lift (PRCL), a score inspired by pointwise $\mathcal{V}$-information that accumulates token-level log-likelihood ratios between a question-conditioned context and a null context with the question removed. This enables prompt scoring from a short rollout prefix rather than a complete trajectory. Across Qwen2.5-Math-1.5B and DeepSeek-R1-Distill-1.5B, prefixes retaining at least 90\% and 95\% of the full-rollout PRCL score, respectively, require only 26\% and 4.7\% of the tokens generated by full rollouts. Compared with SHIFT, LIFT reduces selection cost by 3.7--9.1$\times$ while achieving competitive downstream RLVR performance, with the highest six-benchmark average in three of six settings.

Can Compact Open-Weight Large Language Models Recover Mobility Events from Incomplete Trajectories? A Controlled Evaluation Using Air Trips

Hsuan-Chih Wang, Xianfeng Terry Yang — Tue, 20 Oct 2026 00:00:00 +0000

94% · Evaluation 94%, Hallucination 48%

Incomplete mobility trajectories can obscure the observations needed to identify important travel events and their endpoints. This study evaluates whether compact, locally deployed large language models (LLMs) can recover missing mobility information from incomplete trajectories, using high-confidence air trips as a controlled test case, and whether more explicit prompts can compensate for the loss of endpoint evidence. Using nationwide location-based services (LBS) data from the United States in January 2022 and airport reference data, thirty high-confidence air trips were converted into three incomplete-trajectory conditions by removing observations near the origin airport, the destination airport, or both. Four prompt variants and two LLMs generated 720 requests, each using a two-stage procedure to identify the flight transition segment and then select the origin and destination airports. Exact transition accuracy is 9.2%, while origin and destination accuracy are 31.4% and 10.7%, respectively. Only 1.9% of requests recover both the rule-identified transition and the reference airport pair. The most explicit prompt produced the highest observed exact transition and exact airport-pair recovery rates. The study introduces a controlled endpoint-removal framework for evaluating recovery of discrete mobility events from incomplete trajectories, allowing failures to be examined by inference stage and helping analysts decide whether recovered information should be retained, independently verified, or left unresolved.

Who Speaks, Who Authorizes, Who Acts? Understanding Trust Boundaries in Multi-Agent LLMs

Junru Zhu, Xin Chen, Hang Lyu, Jiayue Meng, Muzi Li, Yifan Chen — Tue, 20 Oct 2026 00:00:00 +0000

94% · Agents 94%, Evaluation 79%

Multi-agent LLM systems exchange messages before acting through tools, but message delivery, signing authority, and tool-side enforcement are separate trust boundaries. We represent them as communication, authorization, and execution structures, and use the authority cut to characterize the smallest independently compromised signer coalition that can construct a policy-valid false-claim proof under stated assumptions. Controlled paired hidden-state studies reveal four distinctions. First, no fixed-message routing contrast across four models meets the prespecified large-effect criterion; when relays rewrite messages, executed false-state action rates vary by model and route configuration. Second, presented source-identity cues do not uniformly increase protected-action selection across three models. Third, on invalid-proof cases, Nemotron and Llama request protected actions in 52.1% and 54.2% of trials, respectively, yet a deterministic tool gate commits none: request-level and commitment-level safety readouts diverge. Finally, a separate single-model authorization check shows that a false claim with a policy-valid compromised-signer proof can pass the gate, since proof validity is not factual truth. These findings motivate evaluating delivered information, qualified authorization, model proposals, and committed actions separately.

Measuring the Task Value of Precision Upgrades in Closed-Loop Agents

Sicheng Lyu, Xinyu Wang, Yu Gu, Lei Ding, Ziyang Song, Xiao-Wen Chang — Tue, 20 Oct 2026 00:00:00 +0000

94% · Agents 94%, Evaluation 89%

A low-precision agent can call its full-precision (BF16) model at selected decisions as a fallback. A learned precision router chooses where to call it from the disagreement between the two models and is judged by end-to-end success. An inserted BF16 action, however, can rescue a failing task or damage a succeeding one. We argue that a fallback's value is the number of tasks it saves minus the number it spoils, measured for the intervention that will actually run. From restored states, we run paired branches that cross which model supplies the current action with which model continues, separating replacement of one action, takeover of later execution, and both. On Qwen3 agents with simulated 4-bit weights in tau2-bench, replacing the first action where the two models differed in tool use changed many task outcomes, but rescues and damages nearly cancelled, and on held-out telecom tasks the action-only estimate was negative. Replacing both the action and the continuation had a positive net gain in two of five exploratory lower-bit stress settings. Two references changed outcomes at fixed bit width: replacing an action with that of a second 4-bit checkpoint changed about as many outcomes as replacing it with the BF16 action, and reordering tool-argument keys changed outcomes while the executed tool call stayed identical. At common states, the additional value of the BF16 action over that checkpoint remained unresolved. A frozen uncertainty trigger found more outcome changes than random selection, while its net benefit stayed within the range of random selection. We recommend evaluating each fallback by its signed task benefit under the intended continuation, against same-precision references, before attributing gains to more bits.

Training with Ciphers to Enhance In-Context Terminology Learning For Translation from Rare Languages

Deevyanshu Malu, Deeptanshu Malu, Sunita Sarawagi — Tue, 20 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 56%

Fine-tuning LLMs with in-context examples (IC-Train) lets translation models adapt to new parallel data at deployment, but it is unclear whether they actually learn the translations of rare terms from those examples. We formalize this capability as in-context term learning (ICTL). We show that standard IC-Train develops this capability only weakly, and propose RareCipherFT, which strengthens it without any lexical resources. RareCipherFT pairs each training target with in-context examples that share a rare term with it. It then replaces that term with a character-shuffled cipher in both the context and the reference, so the term can only be recovered by copying from context. The ciphered fraction of training data is annealed to preserve translation quality. Across seven translation directions between English and Tamil, Assamese, Mizo, Nepali and Hausa, and four models from 1B to 7B parameters, RareCipherFT raises term recall over standard IC-Train by 5-15 absolute points without lowering COMET. Probes using synonyms, nonce words and controls for shared surface context show that these gains come from linking source terms to their in-context translations rather than from shortcuts.

ECHO: Efficient Causal Self-Recursion Beyond Auxiliary Multi-Token Prediction

Tue, 20 Oct 2026 00:00:00 +0000

94% · New model 94%, Efficiency 82%

Existing Multi-token prediction (MTP) methods are auxiliary by design: the multi-horizon loss mostly trains modules attached to the backbone, and each added horizon costs another module and another pass. We propose Efficient Causal Self-Recursion (ECHO), which removes both costs. The backbone's own final layer predicts every horizon, so the supervision lands on the parameters that generate the next token, and Dynamic Depth Sampling (DDS) keeps deeper horizons affordable by drawing the training depth with probabilities that decay along depth. Future prediction thus becomes part of the autoregressive objective rather than a task attached to it, and the gains follow on both sides of it. In pretraining, added depth converts into downstream accuracy where auxiliary modules saturate, concentrating on reasoning and code at half the MTP training overhead of training every depth. At inference, the pretrained model is its own drafter: it drafts beyond the depth it was trained for, and keeps its draft-target alignment on the open-ended, high-entropy generation where post-hoc drafters lose theirs. The advantage survives a doubled token budget, a repeated pretraining run, and supervised finetuning.

DialogMem: Evaluating Dialogue-Grounded Embodied Memory in 3D Environments

SeongJun Jeong, Junghyun Kim, Gi-Cheon Kang, Byoung-Tak Zhang — Tue, 20 Oct 2026 00:00:00 +0000

94% · Evaluation 94%

Embodied agents in human-centered environments require memory that integrates what they observe with what humans tell them over time. However, existing embodied memories are built primarily from visual experience, leaving open how dialogue with humans can be grounded in the objects encountered throughout that experience. We introduce DialogMem, a two-phase benchmark in which agents build memory from time-aligned streams of posed RGB-D observations and dialogue, and later use it for instance-level object navigation and question answering (QA). DialogMem provides 142 episodes across 27 real-world 3D scenes, with 14,193 QA pairs and 11,890 navigation instructions. We evaluate nine memory representations within a unified framework of memory formation, evidence retrieval, and reasoning for task outputs. In experiments, appending dialogue to visual memory at evaluation time substantially improves QA but yields negligible navigation gains, suggesting that relevant dialogue is present in memory but not grounded to specific objects. Controlled diagnostics show that navigation improves with link quality but falls short of ground-truth object-level memory even under perfect linking. Retrieving the correct object entry and recovering its spatial location remain further bottlenecks. These results identify three bottlenecks for dialogue-grounded embodied memory: grounding and organizing dialogue at object-level entries during formation, retrieving the correct entry, and localizing it accurately.

FlowMaker: From Tasks to Cost-Efficient Generative AI Workflow Deployments

Dimitrios Liakopoulos, Prasoon Sinha, Neeraja J. Yadwadkar — Tue, 20 Oct 2026 00:00:00 +0000

94% · Efficiency 94%

Generative AI workflows execute user tasks by composing sequences of LLM calls, where each call performs a sub-task (e.g., summarization or translation), forming a workflow. Executing such workflows requires decisions over workflow structure, model assignments, and deployment configurations, which jointly determine latency, cost, and accuracy under dynamic workloads and resource constraints. Existing systems optimize these decisions in isolation or fix them early, leading to suboptimal cost and goodput. We make two key observations. First, efficient serving requires jointly exploring all combinations of workflows, model choices, and deployment configurations under dynamic cluster state. Second, reuse of deployed LLMs is critical for reducing cost, but must be performed on-the-fly, as reuse depends on current load and resource availability. Together, these make selecting a workflow-variant-a combination of workflow, models, and deployment configuration-a time-critical problem. We present FlowMaker, which exposes a task-level interface and selects workflow-variants on behalf of users. To realize this, we design a constraint-guided load-aware pruning algorithm for fast, on-the-fly decisions. FlowMaker improves goodput by up to 10.14$\times$ and reduces cost by up to 2.77$\times$ compared to state-of-the-art baselines, including Murakkab.

HybridResearcher: Bridging the Structured-Unstructured Boundary for Deep Research Agents

Soyoung Yoon, Peiran Xu, Han Wang, Ruofan Wu, Xiaodong Yu, Feng Yan — Tue, 20 Oct 2026 00:00:00 +0000

94% · New model 94%, Agents 90%

While LLM agents are increasingly capable of gathering and synthesizing evidence from a single homogeneous source, such as structured tables or unstructured data, real-world analytical questions frequently span heterogeneous sources. For example, answering one enterprise question may require both a result computed over an internal database and evidence found on the web. Recent evaluations on benchmarks like HybridDeepResearch show that even frontier agents struggle when evidence must cross this boundary. While these benchmarks have exposed the limitation, methodologies to explicitly teach cross-source reasoning remain absent. To address this, we introduce HybridResearcher, a 4B agent trained to bridge evidence across the structured and unstructured boundary. To build it, we synthesize cross-source bridge trajectories by starting with validated single-source tasks-such as existing text-to-SQL or web search queries-and fabricating the missing source around a bridge entity drawn from the seed's own output. We retain a sample only when strict construction gates confirm that neither side alone suffices, ensuring the data requires true cross-source interdependence. Finally, we train on this curated dataset using supervised fine-tuning followed by reinforcement learning. Evaluated on the existing HybridDeepResearch benchmark, HybridResearcher improves over our base model, Qwen3.5-4B, by 30.0 points pass@8 and surpasses the far larger Qwen3.5-397B-A17B by 8.3 points, while also improving over its base model on established, held-out single-source benchmarks (BrowseComp and BrowseComp-Plus), demonstrating that this complex cross-source reasoning can be taught directly to small models without triggering catastrophic forgetting.

T-MUON: Tensorizing Muon via t-SVD for Large Language Model Pre-training

Dongyang Li, Mingsong Yan, Zheng Zhang, Sui Tang — Tue, 20 Oct 2026 00:00:00 +0000

94% · New model 94%

Muon has recently emerged as a strong optimizer for large language model pretraining by exploiting matrix structure through polar orthogonalization. However, it processes each parameter matrix independently, whereas recent tensor-aware extensions of Muon jointly update groups of eligible matrix parameters through matrix unfoldings. In contrast, we propose T-MUON, a t-SVD-based generalization that groups structurally related Transformer gradients into third-order tensors and defines the corresponding Muon-style update directly through the $M$-transform. We show that the exact T-MUON update is a steepest-descent direction under the $\star_M$-spectral norm and establish a convergence guarantee. Across GPT-2-style and LLaMA-style models from 51M to 1.3B parameters trained on FineWeb and FineWeb-Edu, head-wise T-MUON with the identity transform consistently achieves lower validation perplexity than AdamW, Muon, and variants using DFT, DCT, or DST transforms. These gains persist in additional experiments with longer context lengths and larger token budgets, although the preferred tensorization mode varies across settings.

CasePlay: Self-Play Reinforcement Learning from Case Reports for Medical Reasoning

Hao Wang, Zihan Wang, Kai Wu, Jian Yao, Yiwen Ye, Yongcan Luo — Tue, 20 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 52%

Reinforcement learning (RL) is a promising way to improve the clinical reasoning abilities of large language models (LLMs), but current pipelines often rely on expert-labeled questions or costly synthetic-data generation. Meanwhile, the medical literature offers a wealth of high-quality unlabeled documents, including clinical case reports that trace the full clinical course from initial presentation to final treatment decisions. We study whether document-grounded self-play can convert these reports into effective RL training signals. We identify two challenges that limit existing self-play methods in this setting: knowledge-point fixation}, where the proposer repeatedly asks about a narrow set of salient clues in a source report, and difficulty-quality mismatch, where appropriately difficult questions may still be poorly formed or shallow, limiting their value for reasoner training. To address these challenges, we propose CasePlay, a self-play framework with two key components: (a) a Knowledge-Conditioned Proposer, which anchors question generation to case-specific knowledge points so that questions draw on a broader range of evidence from the source report; and (b) a Rubric-Judge Reasoner, which augments answer-accuracy feedback with rubric-based quality feedback on clue integration, reasoning depth, and option quality, guiding the proposer toward reasoning-intensive questions that better support reasoner training. Extensive experiments across seven medically related benchmarks demonstrate that CasePlay outperforms existing self-play baselines and improves the SFT warm-start model by 3.1 points on average, highlighted by a 6.3-point gain on LiveClin. This approach offers an effective path toward continually improving medical reasoning ability through self-play RL.

MemPrice: Online Memory Allocation via Adaptive Pricing for Long-Horizon Agents

Yiming Zong, Shiyuan Piao, Jiashuo Jiang, Fugee Tsung — Tue, 20 Oct 2026 00:00:00 +0000

94% · Evaluation 94%, Agents 73%, Efficiency 35%

Long-horizon LLM agents face a fundamental allocation problem as interaction histories grow: which information to retain, how much context to devote to it, and at what granularity. Existing memory mechanisms largely focus on relevance, retrieval, or compression, leaving this allocation trade-off implicit. We introduce MemPrice, which treats active context as a shared resource and dynamically allocates it across memories and representation levels. MemPrice turns accumulated memory pressure into an adaptive price on context, encouraging detail only when its estimated retention value justifies the additional footprint. This pressure persists across memory refreshes, allowing the contents of memory to evolve while preserving long-run control over context usage. We show that irreversible compression can force linear worst-case regret against hindsight following its own memory trajectory, and introduce a same-state benchmark to evaluate allocation on MemPrice's realized memory states. MemPrice enforces the hard context budget at every round and regulates long-run occupancy through adaptive pricing, without prior knowledge of the interaction horizon. Under priced-frontier regularity, its average allocation-regret bound against this benchmark vanishes when accumulated estimation error is sublinear. We evaluate the framework under matched memory budgets on LongMemEval, LoCoMo, AppWorld, and WebShop across three foundation models, spanning both query-blind memory construction and goal-directed agent context management.

SteadyThought: Mitigating LLM Under-Thinking via Thought-Level Preference Optimization

Zhenyue Peng, Lin Yang, Jie Wang, Xiaoqi Ni, Hanzhu Chen, Xize Liang — Tue, 20 Oct 2026 00:00:00 +0000

94% · New model 94%, Efficiency 93%, Evaluation 30%

Flexible switching between reasoning trajectories (i.e., thought switching) has significantly enhanced the reasoning capabilities of Large Reasoning Models (LRMs). However, existing models often switch excessively yet fail to sustain promising reasoning thoughts, a phenomenon termed "under-thinking". While recent efforts suppress switching to mitigate this, such over-correction may discard valuable trajectories. To address this challenge, we propose Steady Thought (ST), a novel thought-level preference optimization framework. ST formalizes under-thinking as a preference issue at switching points, forcing single-path continuations from identified points to yield high-quality reasoning. Then, ST performs thought-level preference optimization by treating the newly generated response as preferred and the original one as dis-preferred. Experiments across multiple models and datasets show that ST effectively reduces token consumption, achieving reductions of up to 63.8%, while maintaining strong reasoning accuracy. Further analysis suggests that ST helps models acquire a generalizable ability to reduce redundant switching across domains and languages. Code is available at https://github.com/MIRALab-USTC/SteadyThought.

Conformal Prediction-Enabled Sequence-level Policy Optimization for Reasoning Language Models

Hui Yang, Wan Tian, Tao Ren, Zhongyi Li, Yijie Peng — Tue, 20 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 35%

Reinforcement learning improves the reasoning capability of large language models through verifiable rewards for mathematical and coding tasks. In token-level methods such as GRPO, clipping acts on individual token ratios while the reward and advantage apply to a complete response. Recent sequence-level methods align these units but still use fixed clipping bands, whose tail probabilities change as the score distribution evolves. We propose Conformal Prediction-Enabled Sequence-level Policy Optimization (CSPO), which adaptively calibrates sequence-level clipping bounds using recent training statistics. The resulting band targets a specified marginal tail budget and adapts to policy drift; calibration conditional on response length requires additional assumptions. We bound the probability that a fresh score falls outside the band by the target miscoverage level plus explicit CDF discrepancy and drift terms. Under additional moment conditions, this bound also controls the objective and gradient error relative to the unclipped GSPO power surrogate. Experiments with the Qwen3-1.7B series show improvements over GSPO on difficult mathematical reasoning benchmarks and ALFWorld, with clipping endpoints that adapt during training.

Assertion and Belief: Moore's Paradox in Language Models

Anshun Asher Zheng, Weili Cao, Fangcong Yin, David Beaver, Junyi Jessy Li — Tue, 20 Oct 2026 00:00:00 +0000

94% · Evaluation 94%

Moore's paradox \citep{moore1942} concerns sentences such as ``It is raining, but I don't believe that it is,'' which may well be true yet cannot be sensibly asserted. The absurdity lies not in the content but in the act, since to assert is to vouch for one's own belief. This makes the paradox a test of whether language models take the speech act into account when computing meaning. We present the first test of Moore's paradox in LLMs, evaluating 12 models on a curated set of Moorean sentences in varied contexts. We find that models largely recognize the infelicity of Moorean sentences, and that the judgement is not tied to a memorized form and tracks the speech act. Using the Rational Speech Act framework (RSA; \citealt{goodmanfrank2016}), we then test whether these judgements arise from a listener reasoning about a speaker. We find the models' own listeners do not reason about their own speakers, and their reasoning traces suggest that they instead recall knowledge about the paradox and apply rules therein. The models thus appear to have learned the paradox as a fact about sentences, not as a requirement on speakers.

Learning to Couple: Context-Adaptive Joint Denoising for Diffusion Language Models

Zilei Shao, Ian Li, Heng Zhao, Kewei Lian, Guy Van den Broeck, Zhe Zeng — Tue, 20 Oct 2026 00:00:00 +0000

94% · New model 94%, Efficiency 78%

Diffusion language models accelerate generation by predicting multiple tokens in parallel, but factorized denoisers treat simultaneous predictions as conditionally independent. Decoding heuristics reduce the resulting errors by controlling which tokens are committed, but leave the underlying distribution factorized. Joint modeling addresses the source of the problem, yet existing methods attach separately learned layers to frozen denoisers, leaving the backbone unadapted to the joint objective and the coupling parameters fixed across contexts. We propose Codpula, a context-parameterized denoiser combining a Transformer with a tractable probabilistic inference layer. Given a partially masked input, the Transformer produces token potentials that shape per-position predictions and compact gates that modulate the layer's coupling parameters, yielding a context-dependent joint distribution while retaining exact conditional inference. We jointly optimize both components under a conditional denoising objective using a dual-flow multiplicative update that preserves valid probabilistic parameters. At inference, Codpula supports standard heuristics and uses conditional queries to adapt commitments to dependencies among proposed tokens without extra Transformer evaluations within a step. Codpula sets new state-of-the-art results across a broad suite of reasoning benchmarks while advancing the accuracy-latency Pareto frontier in mathematical reasoning and code generation.

Which Answers Agree? Evaluating Self-Consistency from Incomplete Logs

Zhengtao Gui, Zexi Chen, Weining Shen — Tue, 20 Oct 2026 00:00:00 +0000

94% · Evaluation 94%

Self-consistency answers a question by generating multiple responses and selecting the most frequent answer. Its evaluation requires both correctness labels and information about which responses agree; correctness labels alone generally determine neither voting accuracy nor the benefit of additional votes. For tasks in which all correct responses map to the same normalized answer, we derive sharp accuracy bounds using classical stochastic majorization, refined by abstention counts and answer-category limits. We also derive sharp bounds on accuracy changes across voting budgets that share a response pool. When answer-category counts are available, we introduce Plurality-U, a count-based dynamic program that exactly averages voting accuracy over all fixed-size subsets, eliminating finite-pool Monte Carlo error. On 1,403,520 stored outputs from recent reasoning models in Scorio Lite and Scorio Trace, recording abstention counts and answer options cuts the width of SuperGPQA accuracy intervals by more than half. For moving from 9 to 17 votes, our bounds reduce interval width by a median of 77.7% on SuperGPQA and 72.3% on 14 GPQA and Adaptive-Consistency pools, compared with a baseline that combines separate accuracy bounds. Exact computation has lower paired median runtime than 2,000-subset Monte Carlo in 14 of 15 tested pool-budget settings. These results clarify what incomplete logs can establish and provide an exact evaluation procedure when answer counts are retained.

Correctness-Dependent Asymmetric Target Distributions for Self-Distillation Policy Optimization

Suwan Yoon, Seongryong Jung, Changhee Lee — Tue, 20 Oct 2026 00:00:00 +0000

94% · New model 94%

Self-distillation policy optimization provides dense token-level supervision for LLM post-training through feedback-conditioned self-teachers. However, directly matching the self-teacher can disrupt successful behaviors on correct rollouts while providing insufficient correction on incorrect ones. We propose Asymmetric Policy Optimization via Self-Distillation (APSD), which constructs correctness-dependent targets by mixing student and self-teacher distributions: interpolation preserves successful behaviors on correct rollouts, while extrapolation beyond the self-teacher strengthens correction on incorrect ones. We analyze APSD through divergence decomposition, reverse-KL barycentric geometry, and gradient analysis, highlighting the importance of gradient flow through student-dependent targets. Experiments under reinforcement learning with verifiable rewards and rich feedback show that APSD achieves strong overall performance against reinforcement-learning and self-distillation baselines and matches SDPO performance with 2.6x fewer training steps on average across Qwen3 model scales.

TempoMoE: Identifying and Steering Fast-Slow Thinking Experts for Efficient Multimodal Reasoning

Xiyue Zhang, Keru Zhou, Ziyi Wang, Bing Hu, Li Wang — Tue, 20 Oct 2026 00:00:00 +0000

94% · Efficiency 94%, Evaluation 36%, New model 30%

Multimodal Mixture-of-Experts (MoE) reasoning models rely on long chain-of-thought (CoT) for complex tasks, yet often overthink easy inputs, increasing token cost and latency, and sometimes degrading accuracy. Existing methods optimize reasoning length during training or impose external control at inference, but leave the underlying reasoning mechanism a black box. We find that fast and slow thinking exhibit distinct expert-routing patterns: holding the input fixed and varying only the prompted reasoning tempo produces stable routing shifts. Suppressing slow-thinking experts shortens the reasoning chain, whereas suppressing fast-thinking experts lengthens it, providing causal evidence that these experts influence reasoning depth. Based on this finding, we propose TempoMoE, a training-free, two-stage framework. It identifies fast- and slow-thinking experts via tempo-contrastive routing analysis and redistributes routing weights within the original top-k set according to sample difficulty. It amplifies fast-thinking experts and suppresses slow-thinking experts on easy inputs and reverses this adjustment on hard ones, enabling difficulty-adaptive control over reasoning depth. We evaluate TempoMoE on Kimi-VL, Qwen3-VL, and ERNIE-4.5-VL across six benchmarks. Compared with Vanilla, TempoMoE consistently improves accuracy while reducing output tokens, with accuracy gains of up to 2.6 points and token reductions of up to 35.4%.

CUA-Lite: An Open Platform for Computer Use Agents

Zhanhui Zhou, Weichen Zhang, Haoran Liu, Lingjie Chen, Tianneng Shi, Kevin Lin — Tue, 20 Oct 2026 00:00:00 +0000

94% · Evaluation 94%, Agents 91%, Efficiency 59%

Open resources for computer use agents (CUAs) -- environments, datasets, and models -- remain scattered: without a standard, these resources cannot be integrated and scaled; additionally, evaluation, RL, and SFT cannot share one infrastructure. We present CUA-Lite, an open platform for developing CUAs, built around three standardized abstractions: (1) Unified environment interface, under which we integrate 12 benchmarks across desktop, browser, and mobile, plus lightweight VM-free sandboxes that replicate OSWorld's desktop at much lower cost, providing over 30,000 verifiable tasks. (2) Unified supervised data format, into which we convert the major public CUA datasets, from GUI grounding and understanding to full rollouts, and record our fresh rollouts from frontier CUAs on these verifiable environments. (3) Model harnesses shared across evaluation, RL, and SFT, one per model and implemented for 14 model families. Each harness runs its model in these computer environments, producing the rollouts that evaluation and RL consume, or renders stored supervised data into the model's own training format for SFT. We release reference scores across the integrated benchmarks, and show that RL and SFT through the same harness improve open-weight models. We release our dataset, code, and models, and the platform is open to contributions.

Safety Mid-Training: Internalizing LLM Safety as a Foundational Capability

Zhengyue Zhao, Liwei Jiang, Yejin Choi, Chaowei Xiao1 — Tue, 20 Oct 2026 00:00:00 +0000

94% · New model 94%, Evaluation 32%

Safety alignment for large language models (LLMs) is primarily performed during post-training, while recent work has also begun incorporating safety data into pre-training. Yet safety can degrade when later training no longer includes safety data, and gains in safety often come at the cost of general capability. This tension suggests that safety and general reasoning should not be treated as separate training goals. We instead view safety as an LLM capability that can be learned through a reasoning structure shared with general tasks. Both require the model to understand intent or task goals, apply relevant knowledge, and reason about potential consequences. We introduce safety mid-training, a stage between pre-training and post-training that jointly develops safety and general reasoning through this shared structure. Given safety and utility prompts, the base model generates structured documents and reasoning traces that are then used for mid-training. Safety mid-training reduces average attack success rate (ASR) by 72% to 81% relative to Qwen3-1.7B, 4B, and 8B base models, while also improving general capability. The benefits persist through subsequent post-training. After general-only SFT, mid-trained models largely retain their safety, achieving lower ASR than the baselines without sacrificing general capability. They also retain safety under general-only GRPO without safety rewards, while subsequent RL reaches a better safety-utility frontier when initialized from the mid-trained models. Together, these results suggest that learning safety and general reasoning within a shared mid-training stage can reduce the safety-utility trade-off and make safety more robust to subsequent training.

Zero-Shot Tabular Learning via Hypergraph Reasoning with LLMs

Geon Lee, Kyungho Kim, Dongwon Choi, Juyeon Kim, Shinhwan Kang, Kijung Shin — Tue, 20 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, New model 38%

Tabular data are prevalent across diverse domains such as healthcare and finance. Zero-shot tabular learning aims to make predictions without labeled data. However, existing LLM-based approaches rely on row-wise prompting, resulting in high inference cost and limited use of inter-row dependencies. In this work, we propose TABHGR+, an efficient relation-aware zero-shot framework for tabular data that leverages LLM-generated rules to label and group rows. Specifically, based on column and task descriptions, we use LLMs to generate rules that assign weak labels to all rows. To refine these noisy weak labels, we construct a hypergraph by grouping semantically related rows using LLM-generated rules and perform hypergraph-based label propagation. Lastly, we further enhance the LLM-generated hypergraph for propagation, which is potentially sub-optimal, by assigning label-adaptive weights to hyperedges. By requiring only a constant number of LLM calls independent of dataset size, TABHGR+ substantially reduces inference cost while achieving higher classification accuracy than zero-shot and strong few-shot baselines.

PeekKV: The First Content Token Already Knows Which Heads Matter

Minyoung Go, Yongjun Park — Tue, 20 Oct 2026 00:00:00 +0000

93% · Efficiency 93%

The size of the KV cache grows linearly with context length, making it a major memory bottleneck in LLM serving. Motivated by recent progress in identifying the functional roles of attention heads, a growing body of work compresses the KV cache non-uniformly across heads. Existing methods either fix the important heads via offline profiling or select them using attention signals from the prefill stage. In both cases, however, accuracy fluctuates substantially across workloads. While investigating the cause of this instability, we find that the active heads shift between the prefill and decode stages even within a single input. Along the decode path, by contrast, the activehead composition remains stable across content tokens once format tokens are filtered out. Based on this observation, we propose PeekKV. PeekKV observes a single content token immediately after prefill - a one-token peek - to jointly obtain head-group scores and token-importance scores. Guided by these scores, PeekKV allocates KV cache budgets to head groups in proportion to their group scores and retains only the top-scoring tokens within each budget. PeekKV requires no offline profiling or training and thus applies directly to unseen workloads and new models. Across four public benchmarks, PeekKV achieves the highest accuracy in more than half of the per-model comparisons against state-of-the-art methods, and stays closest to the full-cache accuracy on RULER and LongBench. On Llama-3.1-8B-Instruct, PeekKV retains 95.7% of the full-cache accuracy at a 10% memory budget on RULER with 16K contexts and 95.9% at a 1.5% budget on LongBench.

Cortex: A Fixed-Point Theory of Governed Coding Agents - Technical Report

Marius-Constantin Dinu, Florian Zeba — Tue, 20 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, Agents 65%

A coding agent combines a large language model (LLM) with a harness that plans, edits, and executes code. We study Cortex, a supervisory layer for pre-execution governance, requirement tracking, and iterative validation and repair. We formalize requirement closure on a finite lattice. For a fixed requirement set and a sound, monotone validation rule with persistent evidence, exhaustive iteration reaches the least fixed point with no more strict increases than there are requirements. Fair consequence scheduling reaches the same closure. A separate stochastic model bounds expected completion time under a uniform positive-progress assumption. These results depend on their stated assumptions; artifact repair, fallible validation, and finite execution budgets do not satisfy them automatically. We define Gauntlet's trajectory-similarity and build-gated scores, attack-success estimators, Wilson intervals, and case-cluster bootstrap summaries. Effective-feedback accounting counts newly certified requirements, rather than estimating their information value or utility. The five evaluation families support raw-versus-governed comparisons. We report retained benchmark observations, including rescoring of existing artifacts; constructed outcomes are excluded. Unequal observation sets and incomplete provenance prevent causal claims about governance or measured long-horizon scaling.

Best of Both Worlds: Fine-Tuned Small Models Complement LLMs for Cost and Performance Benefits

Tue, 20 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 36%

BERT-based models require thousands of labeled samples to train. Larger models like GPT-X require only 1-5 examples per class. However, performance, cost and time tradeoffs of such models remain under-explored. In this work, we classify survey text data into 16 classes. We focus on experimental approaches and results on performance, cost, and time taken for this problem. We compare these parameters of a trained RoBERTa-based model, and an (off-the-shelf) GPT-3.5 model for text classification. Baseline experiments show that the time taken by GPT-3.5 for generating predictions on 10,000 records is 6 hours, whereas a RoBERTa-based model takes only 12 mins on a T4 GPU. To identify the best prompting approach for GPT-3.5, we explore batch prompting. We experiment with (1) batch-size, and (2) few-shot sampling. We find the best batch size is 20 with human few-shot curation. Experiments show the F1 score of a trained RoBERTa-based model (96%) is much higher than a GPT-3.5 model (83%). In case of GPT-3.5, batch prompting is beneficial in lowering inference costs by ~70%, without hindering task performance. We find the best approach is an ensemble of a RoBERTa-based and GPT-3.5 model (F1: 97.5%) incurring only 4.4% of inference cost compared to a standalone GPT-3.5.

Axiomatic World Modeling for Physics Reasoning

Xinye Yang, Zhenyang Liu, Yuxuan Wang, Yuanyuan Lei — Tue, 20 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 74%, Hallucination 67%

Recent research-level physics-reasoning benchmarks for LLMs reveal a failure mode that lies not in knowledge or execution, but in modeling the world being reasoned about. Models often silently alter given quantities, fabricate laws, or inject unstated assumptions to force a problem into familiar patterns, and then reason inside this corrupted world. We define this premise-level divergence as modeling drift: the gap between a model's internal axiomatic world model (AWM) and the true physical world specified in the problem. Although RLVR and PRM have yielded substantial gains in mathematics and code, their reward shapes do not directly supervise world modeling in axiomatic domains such as physics. We propose Reinforcement Learning with Axiomatic World Modeling (RLAWM), which optimizes the policy by penalizing divergence between its inferred axiomatic world model and the true physics world. RLAWM probes physics reasoning at two hierarchical levels (modeling and solving), each scored by a physically grounded reward. The modeling reward targets abstract knowledge formulation, evaluating whether the model captures the true axiomatic world before attempting calculation. The solving reward targets instantiated knowledge reasoning, checking whether the model's concrete derivation remains consistent with the axiomatic world. This two-level structure separates what world the model believes it is reasoning in from how it reasons within that world. The framework operates exclusively during training, with no extra cost during inference. RLAWM improves performance by up to +12.2 points on PhysReason and +5.2 on PHYSICS over the baseline. It also generalizes zero-shot across difficulty, source, and domain, with a 2.45x gain and +10 points over the strongest baseline on olympiad-level physics and math.

Neuro-Inspired Inhibitory Message Passing for Accurate and Token-Efficient Multi-Agent Reasoning

Nima H. Siboni, Magdalena Springer, Vahid Rostami — Tue, 20 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Agents 41%

Multi-agent systems promise to solve problems by bringing together information beyond the reach of any individual agent. Retrieval and tool use can distribute this information across agents, making communication essential to collaboration. A central challenge is keeping crucial evidence visible in each agent's context as messages accumulate. Repeated information can crowd out rare, decisive evidence. Models of cortical circuits suggest a principle for addressing this imbalance: inhibition can selectively suppress shared inputs while preserving sensitivity to weaker signals. We adapt this principle into inhibitory message passing, a training-free communication router that cancels redundant evidence separately for each evidence sentence and receiver. Each round, it forwards evidence sentences restated in agents' messages, delivering each at most once to receivers whose own evidence lacks it and excluding the senders' reasoning and answers. Against standard message passing, accuracy in peer groups rose by 39-77 percentage points across three benchmarks, and tokens per run fell by 10-46%. Gains held across seven open models spanning four to thirty-two billion parameters and two commercial models, in groups up to sixty-four agents, and across multi-agent architectures including peer-to-peer, orchestrator-worker, and multilevel hierarchical organizations. These findings support a scalable design pattern for more accurate and token-efficient multi-agent reasoning.

PV-Decoding: Progressive Verification Decoding with Decode Gating for Diffusion Language Models

Renxi Liu, Yiming Li, Shengyin Sun, Chen Chen, Xian Wang, Shuai Yang — Tue, 20 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 38%

Discrete diffusion language models (dLLMs) offer an attractive alternative to autoregressive generation by updating multiple tokens in parallel, but their inference efficiency is still limited by the cost of safely committing tentative predictions. In particular, speculative decoding for dLLMs is challenging because speculative tokens within the same diffusion block may depend on one another, making aggressive acceptance prone to error propagation and strict verification computationally expensive. We present PV-Decoding, a confidence-driven speculative decoding framework with progressive verification for block diffusion inference. PV-Decoding partitions masked positions into directly decoded tokens, speculative tokens, and unresolved tokens using two confidence thresholds, and then verifies speculative tokens with progressively stronger policies, including single-batch, dual-batch, and multi-batch verification. This design exposes a practical accuracy--efficiency trade-off: weaker verification minimizes overhead, whereas stronger verification improves acceptance precision and robustness. Experiments on three diffusion language models across GSM8K, MBPP, and HumanEval show that PV-Decoding reduces sequential model forward passes (MFPs) by 8%--37% relative to confidence-based decoding, while keeping the largest observed accuracy drop to 1.3%. On Dream-7B-Instruct/GSM8K, PV-Decoding (1-batch) and PV-Decoding (2-batch) improve throughput by 7.0%--13.6% under the same hardware budget, whereas PV-Decoding (multi-batch) is slower under the current runtime. These results suggest that confidence-guided speculation with structured verification is a practical path toward efficient and reliable dLLM inference.

What Can't Jev Decide? Probing the Limits of One-Pass Decision Models

Wang Yang, Xinpeng Li, Yanyan Zhang, Chuang Ma, Shouren Wang, Jianzhe Wei — Tue, 20 Oct 2026 00:00:00 +0000

93% · Evaluation 93%

Jev and its open replicas choose among the options a caller supplies with a single forward pass, fast enough to decide inside an agent loop. We ask what such a model cannot decide, and find a simple limit: Jev cannot make a decision that an LLM makes only by writing out intermediate steps. On a benchmark that varies one factor at a time, Jev finds and selects records without error up to 20k tokens and 64 options, but once the answer depends on two or more updates of one user, accuracy falls, to 53-80% at sixteen updates, because similar records interfere. Given a stale summary of the state, Jev follows it. On real decision items graded by how many thinking tokens a reference LLM needs, Jev answers 97.5% of the items that one forward pass of that LLM solves, but only 74% of the others. Jev also does not know its limit: its confidence flags the errors it makes itself but not those it inherits from a stale input, and neither an explicit nor a trained hand-off option catches the latter. The open models we evaluate share these limits. Checking the state such models rely on requires a model that reasons.

Certified Sparse Attention via Perturbed Retrieval

Runguo Li — Tue, 20 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 51%

Sparse attention speeds up long-context decoding by reading few keys of the KV cache, but existing methods do not control the error of a query, and on real KV caches their errors have heavy tails. We show that perturbing the score of every key with an independent Exp(1) variable turns threshold retrieval into sampling keys with probability proportional to their attention weights. Lowering the threshold along a geometric grid gives nested samples, a Horvitz--Thompson estimate of the output at every level, and confidence bounds that hold simultaneously over all levels, so each head stops at the first level whose certificate meets the target error $\varepsilon$. With fresh perturbations at every decoding step, each certificate fails with probability at most $\delta$ during generation. At a suitable threshold, the estimator reads $O(\varepsilon^{-2}\log(1/\delta))$ values regardless of the context length, and no adaptive method reads fewer, up to constants. A 4-bit scan with deterministic bounds, shared across a grouped-query-attention group, generates the candidates exactly but reads every key. On three LLMs and ten settings at 16K--128K tokens, no certificate is violated, and at equal keys read our 90th-percentile error is 4.6--7.9$\times$ lower than that of the best of 15 baseline families. At equal bytes, the best baselines have 0.6--1.6 times our 90th-percentile error and at least 4.9 times our largest. End to end, our method stays within one point of dense attention on tasks and within 0.1\% in PG-19 perplexity, and the strongest baselines at equal bytes score as well. On CPU, decoding steps at 32K--128K tokens run 1.1--1.4$\times$ faster than dense attention on the same 16-bit cache.

Legal Guidance in the Wild: How Users Seek Legal Help in Real-World LLM Conversations

Amirreza Mohseni, Gerasimos Spanakis — Tue, 20 Oct 2026 00:00:00 +0000

93% · Evaluation 93%

People ask conversational AI models about their legal rights, obligations, procedures, and personal situations that may have legal consequences. To study this behavior, we first need to distinguish requests for legal guidance from other legally themed discussions and unrelated conversations. We contribute a multi-axis taxonomy of legal queries capturing whether legal guidance is sought, how a request is formulated, and which legal domains it concerns; apply it to 2,122 naturally occurring WildChat-1M conversations; and create a human-reviewed gold evaluation set of 200 conversations. All annotations, model checkpoints, and evaluation code are publicly available. In this enriched sample, 71.1% of conversations containing legal content seek personalized or practical guidance. Rights-and-options advice is the most common primary request type among legal-content conversations, and 25.3% of guidance-positive conversations span multiple legal topics. The gold evaluation set supports a two-stage comparison of ModernBERT-base and ModernBERT-large against a TF-IDF cascade. Our framework supports large-scale studies of legal guidance in conversational AI, and the public release of the annotated data facilitates reproducible research and future model development.

Indistinguishable Agents: Training Agent Privacy as Twin-World Indistinguishability

Jiaming He, Guanyu Hou, Yi Yu, Xudong Jiang — Tue, 20 Oct 2026 00:00:00 +0000

93% · New model 93%, Agents 89%, Evaluation 70%

LLM agents search, shop and send messages for their users, and every action is seen by some party. Existing privacy training rewards agents for not mentioning protected items, yet an observer can infer a private fact from what the agent does. Since an observer learns a fact only if its view changes with the fact, we propose \methodfull (\method), which trains on twin worlds that differ only in the private fact. A frozen distinguisher scores each observer's view by its log-likelihood ratio for the true world, and paired GRPO trains the agent on task success minus this ratio. We prove that this objective controls the Bayes success of every observer on every channel. On shopping and messaging tasks \method removes the leakage that disclosure judges and query-level rewards miss, while the agent still completes its tasks.

EG-GRPO: Evidence-Guided Step-level Credit Assignment for Deep Search Agents

Sijia Luo, Mengjie Gao, Yuxiao He, Mengshu Sun, Cuiping Li, Hong Chen — Tue, 20 Oct 2026 00:00:00 +0000

93% · Agents 93%, New model 92%, Evaluation 50%

LLM-based search agents demonstrate strong capabilities in long-horizon information seeking and reasoning, typically trained via reinforcement learning. However, existing approaches predominantly rely on trajectory-level outcome rewards, providing limited supervision for complex search and reasoning processes. Although step-level credit assignment offers a promising solution, obtaining reliable and task-grounded supervision for open-ended search remains challenging. In this paper, we propose Evidence-Guided Group Relative Policy Optimization (EG-GRPO), a framework that derives step-level supervision from the evidence annotations available in synthetic deep search data. EG-GRPO identifies the first supporting citation for each required evidence claim and uses the newly acquired evidence to construct step-level process rewards. By representing the accumulated evidence as an evolving knowledge state, EG-GRPO further performs state-conditioned relative advantage estimation and combines the resulting step-level signal with the standard trajectory-level outcome advantage. Experiments on five challenging deep search benchmarks show that EG-GRPO generally outperforms standard GRPO and existing reward-enhanced and step-level RL baselines across different model scales. These results demonstrate that task-grounded evidence annotations can provide effective process supervision for credit assignment in deep search RL.

System X: An LLM Decision Model That Replies from Its Own KV Cache

Syed Usama Khalid Bukhari, Syed Safwan Khalid, Muhammad Aamir — Tue, 20 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, New model 71%, Evaluation 61%

Decision models return probabilities over supplied options, but a deployed agent may also need to reply in words. The present study introduces System X, a large language model (LLM) decision model that adds a text-reply option to the LLM-as-Jev readout. When this option wins, or confidence is low, the same model continues from the decision's key-value (KV) cache, avoiding a repeated prompt prefill. Qwen3-0.6B has been fine-tuned with LoRA on Reason-9k and Bitext reply rows. JevBench accuracy shows no statistically significant change (55.4% to 54.5%, p=0.87), while expected calibration error falls from 0.292 to 0.169. Held-out Bitext routing accuracy reaches 98.3%. Cache reuse saves 15-37% of latency on an Apple M4 Pro, with token-identical replies in the FP32 demonstration cases. Scoring options on copies of the prompt cache matches full-sequence scoring to within 1.4e-6 in probability. Adding dialogue, industry and complaint data (System X+) improves support-task results but lowers IFEval accuracy from 55.5% to 45.0%. System X v4 uses Qwen3-1.7B, rebalanced data and teacher-rewritten reply targets. Against its base LLM, JevBench accuracy rises from 55.0% to 61.0% (p=0.04), with no statistically significant IFEval change (p=0.71). A matched latency test on unadapted Qwen3-1.7B saves 17-39%. One LLM judge prefers v4 replies to an unadapted 1.7B second call (31 to 11 pairs, p=0.003); the other's preference is not significant. These reply comparisons also differ in prompts and generation limits. State conflicts and unsupported reply details remain failure modes. Further studies with several seeds, human raters and real support data are recommended.

Self-Distillation for Supervised Fine-Tuning

Wenxuan Zhu, Bang An, Shuming Liu, Wenqing Cui, Tianqi Xu, Bing Li — Tue, 20 Oct 2026 00:00:00 +0000

93% · New model 93%

Supervised fine-tuning treats the expert demonstration or gold response as a one-hot target, suppressing valid alternative continuations that the model would otherwise assign probability to. We propose Gold-Conditioned Distillation (GCD), a method that turns expert demonstrations into model-native supervision. GCD uses one model as both teacher and student: the teacher conditions on a prompt containing the gold response in context, the student conditions on the original prompt as at inference, and the student is trained to minimize the forward KL divergence to the teacher's next-token distribution at each gold response token. This lets the model use its prior knowledge to teach itself the new task: correctness is anchored by the gold response, while the soft target preserves plausible alternatives. Across various training tasks on Qwen and Llama models, GCD improves new-task accuracy over SFT and generally reduces degradation on general benchmarks.

Zephyrus-R1: Agentic Reinforcement Learning for Weather Reporting

Tue, 20 Oct 2026 00:00:00 +0000

93% · New model 93%, Agents 92%, Evaluation 84%

Reinforcement learning (RL) has emerged as a key method for training AI models in scientific applications, but RL remains difficult when outputs cannot easily be verified against a known ground-truth answer. Natural-language weather forecast reports are one such domain: they translate atmospheric data into assessments of weather evolution, hazards, and uncertainty for both the general public and key decision-makers. Unlike numerical weather-analysis questions, they admit many valid outputs, which limits deterministic verification. To address this, we develop Zephyrus-R1, a practical framework for RL on scientific weather forecast reporting. We focus on three key components of the agentic RL pipeline: reward design, curriculum learning, and an analysis of human-expert alignment. To encourage both report legibility and factual accuracy, we design a reward function that combines a meteorological rubric with bidirectional claim entailment, both evaluated against National Weather Service reference reports. Our framework optimizes this reward in a simulated tool-calling environment, enabling vision-language agents to gather and analyze forecast evidence across regions and lead times before producing a report. We use annotations from three meteorologists to audit the claim labels produced by our LLM judges. Finally, a curriculum helps small models first develop tool use and scientific reasoning through RL on verifiable weather-analysis questions, then optimize the open-ended report reward. After training Qwen3.5-4B with Zephyrus-R1, we improve the mean across six judge-metric scores by 26% over the base model and achieve results competitive with frontier open-weight models. For Qwen3.8-27B, we improve the same mean by 13% over the base model, by 6.7% over Kimi-K3, and by 2.7% over GPT-5.6-Luna.

Whose Credential Is It? How Owner Attribution in Agent MemoryShapes Cross-User Account Misuse

Hau-Sing So, Daixian Liu, Carl Hou, Yuhaohang He, Mengzhu Wang, Hai-Tao Zheng — Tue, 20 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, Agents 73%

Shared long-term memory lets an LLM agent serving one user inadvertently retrieve another user's credentials, but it is unclear when such exposure leads to actual harm. We introduce a three-stage audit that separates three events: a foreign credential reaches the prompt (exposure), the agent acts with it (behavior), and the other user's account changes (state). In AppWorld, a store that mixes 763 past logins from all users retrieves another user's credential for 84.7 percent of required apps, yet agents rarely act on it. Misuse appears in a stress test where memory lacks the user's own credential and presents a single foreign one as the user's record: seven agents then use it in 80 to 94 percent of runs and change the victim's account in up to 43 percent, sometimes passing the official check. Naming the owner before or inside each credential line prevents most of this: with all else fixed, misuse falls from 92 to 1 of 305 paired runs, also with name-free usernames, though owner-bound notes cost 3 to 5 successful tasks per model on clean memory. A wrongly written owner largely restores misuse; trailing owner tags and added safety instructions help only some models. Agent memory should consistently record whose credential it stores, and evaluations should measure behavior, not only exposure.

PFedQFT: Personalized Federated Quantum-Inspired Fine-Tuning for Medical Large Language Models

Qihan Huang, Lingxia Zhang, Guye Zhang, Zizhu Wang, Xiaowei Mao — Tue, 20 Oct 2026 00:00:00 +0000

93% · New model 93%, Efficiency 65%, Evaluation 30%

Medical data remain fragmented across institutional silos because of privacy, governance, and ownership constraints. Federated learning enables medical institutions to adapt large language models (LLMs) collaboratively without transferring raw records. However, independently parameterized low-rank adapters incur repeated parameter and communication costs across Transformer projections, while uniform aggregation may limit adaptation to heterogeneous institutional data. We propose PFedQFT, a quantum-inspired framework for personalized federated fine-tuning with a compact shared adapter. Specifically, PFedQFT constructs reusable orthonormal adaptation bases from compact skew-symmetric generators using a truncated matrix-exponential approximation followed by QR orthonormalization. Compatible projections reuse these bases, while a globally shared full core preserves the flexibility of each projection. Each client further maintains a lightweight private residual operator, whose directions are learned globally and whose coefficients remain local. Experiments on seven public medical tasks and multiple LLM backbones show substantial reductions in trainable parameters and per-round adapter upload volume, with competitive predictive performance on most tasks.

Does Unlearning Survive Collaboration? Robust Unlearning for Shared-Model LLM Teams

Hengsong Liu, Liangyu Teng, Jingnan Cai, Yuntian Shi, Jingyu Zhang, Yang Liu — Tue, 20 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 77%

LLM unlearning is essential for removing sensitive or unwanted knowledge while preserving useful capabilities. As LLM teams are increasingly used to solve complex problems, forgetting must remain effective during collaboration. Despite advances in forgetting effectiveness, efficiency, and robustness to relearning, team deployment remains underexplored. We find that collaboration weakens forgetting: final-answer evaluations miss recoverable evidence in team messages, including complementary clues supporting joint recovery. We establish theoretically and confirm empirically that controlling final answers or individual messages does not ensure low recovery from their combination. To address this challenge, we propose Aggregation-Conditioned Unlearning (ACU). ACU trains a shared model using reconstruction feedback from individual and pooled messages, prioritizes views with higher recovery risk, and uses replacement-based credit assignment while protecting direct and collaborative utility. TOFU evaluates trajectory-level recovery, while WMDP-bio tests native forgetting and retained utility. On TOFU, ACU lowers full-trajectory recovery from 41.8% to 19.3% in ROUGE-L recall and factual recovery from 34.2% to 18.7% versus output-supervised training, with comparable retained-task performance. These results demonstrate that ACU curbs information leakage in LLM teams and generalizes across team sizes and readers.

Truncated Reasoning Under a Budget: Evaluating System-1 Models as Decision-makers

Zhongyin Wang, Zhaoqian Xue, Junting Mu, Haozheng Liu, Ye Yuan, Haolun Wu — Tue, 20 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 88%, New model 44%

Large language models (LLMs) often generate more reasoning tokens than needed. Limiting the reasoning budget reduces cost, but forcing an answer from an incomplete trace can hurt accuracy. System-1 decision models select among candidate options without producing a reasoning response. We evaluate Jev as a decision reader of truncated reasoning across seven LLM reasoners and five multiple-choice benchmarks (35 pairs). Jev achieves higher mean accuracy than forced answering at tested budgets from 500 to 16,000 tokens, including the estimated costs of reasoning and answering. At a 1,500-token budget, it improves accuracy by 4.1 pp overall and 8.6 pp on truncated questions, with significant gains in 17 pairs under Holm-corrected bootstrap tests. Offline replay estimates 34.5% of the full-run cost, with a 1.13 pp accuracy loss. A policy selected on training folds saves 71.3% of the estimated full-run cost on held-out folds, with a 0.58 pp accuracy loss. We also find that Jev shows strong question-only accuracy, exceeding 500-token forced answering on average (68.0% vs. 65.5%), although this advantage varies by benchmark. These findings suggest that System-1 decision models can complement LLM reasoning and reduce the need for costly reasoning completion.

From Trace to Fix: Residual Task State in Long-Horizon Agents

Zile Huang, Ser-nam Lim — Tue, 20 Oct 2026 00:00:00 +0000

93% · Evaluation 93%, Agents 65%

Recovering from failures in long-horizon, stateful language agents is challenging because the action introducing a defect, the symptom exposing it, and the persistent state requiring repair frequently diverge. Existing methods largely focus on retrospective failure attribution, leaving sequential diagnosis and state repair poorly integrated under constrained budgets. We introduce TRACE-FIX, a unified recovery framework centered on versioned residual task state. Trace transforms execution logs into a structured dependency-linked ledger, prioritizing unresolved task requirements rather than salient surface errors. Fix formulates recovery as budgeted planning over admissible actions, incorporating outcome forecasting and explicit repair-reserve constraints to prevent premature budget exhaustion. To rigorously decouple localization accuracy from downstream repair capability, we present LongH-Debug, evaluating root-cause attribution, verified-fault recovery, and end-to-end closed-loop execution independently. Our empirical analysis demonstrates that structured residual tracking substantially improves causal localization over raw-trace baselines, while budget-aware sequential planning converts verified fault signals into more reliable task recovery without destructive trial-and-error. Code is available at https://github.com/toffeecat/tracedp.

Not Every Rollout Needs Distillation: Correctness-Filtered On-Policy Self-Distillation

Harrison Zhao, Lei Huang, Fei Liu, Wen Shi, Zijian Jin — Tue, 20 Oct 2026 00:00:00 +0000

93% · New model 93%, Efficiency 84%, Evaluation 30%

On-policy self-distillation (OPSD) supervises student-generated trajectories with a reference-conditioned version of the same base model. We ask whether every trajectory supplies distinct supervision that warrants its distillation cost. Gradient measurements on Qwen3-1.7B at initialization reveal that correct and incorrect responses have similar individual gradient magnitudes, but incorrect responses have approximately three times the coherent gradient share. Their estimated population-mean directions also largely overlap. A signal--noise model fitted at a small group size tracks alignment at larger group sizes without refitting. Motivated by these measurements, we introduce correctness-filtered OPSD (C-OPSD), which generates one response per problem and distills only a subset of incorrect responses. Removing unselected rows before both distillation forwards avoids their teacher/student forward and student backward work. Across Qwen3-1.7B, 4B, and 8B on AIME24, AIME25, and HMMT25, C-OPSD achieves on-par performance with full OPSD. In a measured micro-batch-size-four implementation, executed rows decrease by approximately 68%, and with retained rows packed into full micro-batches, training takes about 40% less time while rollout generation remains unchanged. Together, the measurements and training results connect outcome-conditioned supervision redundancy to more compute-efficient distillation.

Learning to Reason Efficiently with Interventional Step Credit

Guanda Li, Xuefeng Bai, Yulong Chen, Zhiyong Wang, Kehai Chen, Min Zhang — Tue, 20 Oct 2026 00:00:00 +0000

93% · New model 93%, Efficiency 92%, Evaluation 69%

Large reasoning models often produce lengthy chain-of-thought (CoT) traces mixing useful intermediate steps with unnecessary ones. Recent methods explore step-level signals such as attention mass and step entropy to identify and suppress unnecessary steps. Despite their effectiveness, such signals reflect the model's behavior at a step rather than that step's contribution to the answer. To address this, we propose the interventional step credit (ISC), which takes a causal-intervention view and assesses step importance to the reasoning through targeted parameter-level interventions. Building on ISC, we further propose IGPO, which uses ISC to guide fine-grained advantage rescaling in reinforcement learning, reducing redundant steps while preserving the important ones. Experiments across two model scales and nine benchmarks show that IGPO achieves up to 62% average token reduction and better overall performance among compared methods.

Asynchronous On-Policy Self-Distillation

Xidong Wu, Qi Qi, Yuqiong Ji, Shangqian Gao — Tue, 20 Oct 2026 00:00:00 +0000

93% · New model 93%, Efficiency 69%, Evaluation 30%

Unlike on-policy distillation (OPD), which depends on a separate teacher model, on-policy self-distillation (OPSD) leverages a single language model acting simultaneously as the student and the teacher, with the teacher conditioned on privileged information. Despite this architectural simplicity, synchronous rollout generation dominates overall training time, creating a severe systems bottleneck when scaling OPSD. To address this, we introduce Asynchronous On-Policy Self-Distillation (AsyncOPSD), a framework that decouples rollout generation from learner updates to improve training throughput. Since the teacher and student share parameters, AsyncOPSD eliminates the multi-model coordination overhead inherent to OPD architectures in asynchronous training. As the student policy evolves, buffered trajectories generated by older policies become stale, potentially degrading performance. To mitigate this, we introduce a distillation objective based on staleness-aware policy control and token-level entropy weighting. Across three model sizes and mathematical reasoning benchmarks, AsyncOPSD improves the training throughput by $1.96$--$2.25\times$ over synchronous OPSD while increasing average benchmark accuracy.

CALM: Consistency-Aware Logit Masking in Decoding Space for Mitigating Overthinking of Large Reasoning Models

Shengjing Liu, Feiyang Li, Yuxuan Yang, Yile Wang, Hui Huang — Tue, 20 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, Evaluation 31%

Large reasoning models (LRMs) have achieved strong performance on challenging mathematical and scientific tasks through long chain-of-thought reasoning, but they often suffer from overthinking with unnecessary reasoning steps. Existing truncation-based methods typically detect consecutive trial-answer agreement and force early exit, yet such local stopping signals do not necessarily indicate final-answer stability and can disrupt useful subsequent reflection or verification, potentially leading to performance degradation. We propose consistency-aware logit masking (CALM), a training-free decoding intervention for mitigating overthinking in LRMs. CALM first monitors trial answers at step boundaries and then applies logit masking to action transition point (ATP) tokens such as "Wait", "Alternatively", and "Hmm", suppressing redundant transitional reasoning while allowing the model to continue its thought process more naturally. Experiments across four advanced LRMs and five benchmarks show that CALM reduces token consumption by 18.6%-25.0% compared with full CoT reasoning while maintaining performance, and can even improve accuracy by 0.07%-1.48% by reducing redundant reasoning. LLM-as-a-Judge results also demonstrate that CALM provides more coherent and concise results.

Claim Laundering: When Agent Decision Models Trust Records Over Reality

Tue, 20 Oct 2026 00:00:00 +0000

93% · Agents 93%, Evaluation 68%

Agent decision models select actions in a single forward pass from inputs that combine environmental observations with the agent's own record of previous actions. When an action fails silently, these sources can conflict: the record reports success, but the environment shows that the step did not take effect. We ask which source determines the next decision. Across Android screen traces, ALFWorld, and WebShop, we construct paired states that independently vary the observation, the logged action, and an explicit success claim. In all 12 configurations of four decision models and three domains, records have a larger average effect on commitment decisions than observations; after claimed-but-failed steps, the models still stop, advance, or purchase in 53-100% of cases. Yet the relevant evidence is often represented internally: linear probes on intermediate hidden states distinguish successful from failed steps on held-out apps, tasks, and goals (AUROC 0.77-1.00), and activation patching reveals increasing influence from records in later layers. We use this signal in EVIGATE, an intermediate-layer veto that requires no additional forward pass and adds 0.04-2.0% serving latency. Under injected silent failures, EVIGATE improves task success by 18.7 points in live ALFWorld agents, 5.0 in WebShop, and 5.1 in a phone simulator, and in WebShop and the simulator it also outperforms random vetoes at the same rate. Its effectiveness depends on representative calibration states: gates calibrated on offline counterfactual states do not transfer reliably to live agent trajectories.

TimesJev: Thinking Fast on Time Series with System 1 Sufficiency

Chenghao Liu, Wei Fan, Jindong Han, Dongjie Wang, Aniket Mahanti, Pengyang Wang — Tue, 20 Oct 2026 00:00:00 +0000

93% · Efficiency 93%, New model 83%

Large language models (LLMs) have recently shown strong potential for time series reasoning, while models that reason over time series are usually with long chains of thought. However, this process often consumes thousands of tokens, directly leading to substantial latency during inference. In contrast, System 1 models such as Jev make decisions instantly and can naturally be an alternative, but raw time series are not compatible with the embedding space of these models, thus hindering the reasoning performance, especially when input series contain thousands of numeric points. In this paper, we make a very early attempt to propose a time series System 1 model, namely TimesJev. Specifically, we first formalize System 1 sufficiency in time series reasoning, which motivates a compressed representation for time series with key frequency patterns preserved. To construct this representation, we derive a dimension reduction objective that combines (i) a sufficiency divergence that measures the core information lost during compression, and (ii) a realizability divergence that quantifies the incompatibility of the System 1 model to the optimal prediction given this representation. We optimize this objective by learning time series compression with a gating mechanism for sufficiency, and then aligning the selected representation with the embedding space for realizability. Finally, experiments on 13 datasets demonstrate that TimesJev outperforms reasoning models by over 62% with 62.9% less inference time than their fine-tuned versions on average.

Nash-MOPD: Optimizing Task Trade-offs in Multi-Teacher On-Policy Distillation

Siyuan Gan, Bo Yang, Linjian Meng, Boyan Wang, Xiaolong Qi, Jing Huo — Tue, 20 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 31%

Multi-teacher on-policy distillation (MOPD) aggregates the capabilities of multiple domain-expert teachers into a single student model. However, all domains in MOPD share the student's finite budget of optimization steps and rollouts. Fixed uses constant domain sampling ratios based on corpus sizes; by contrast, D$^3$-MOPD adjusts the ratios using the teacher-student reverse-KL divergence and its recent decrease. However, favoring large recent decreases can direct more training toward domains that already improve quickly, while domains that improve more slowly may receive too few examples. To address inflexible allocation and the risk of giving insufficient training to domains that improve more slowly, we propose Nash-MOPD, an adaptive domain scheduler based on a local minimax game using existing reverse-KL statistics. The equilibrium maximizes the smallest predicted proportional reduction under the local model and gives more prompts to domains that need more examples for comparable improvement. Consequently, Nash-MOPD responds to changing learning progress and supports domains that improve more slowly without extra teacher queries or rollouts. At step 100, Nash-MOPD exceeds both baselines' final mathematics and code checkpoint scores on DeepSeek-R1-Distill-Qwen-1.5B and reaches Fixed's on Qwen3-8B-Base. Furthermore, it leads eight of nine reported metrics and is best or tied best on all six mathematics and code accuracy metrics, respectively.

Bridged Self-Distillation for On-Policy Learning

Tue, 20 Oct 2026 00:00:00 +0000

93% · New model 93%

On-policy distillation can transfer the reasoning capabilities of a strong teacher to a smaller student. However, distillation from a stronger teacher model faces two mismatches: its solutions may be difficult for a smaller student to imitate, and its token-level probabilities may be incompatible with, or even unavailable to, the student. Existing methods handle these settings through different forms of supervision. To this end, we propose Bridged Self-Distillation (BSD), a unified framework for distilling from any teacher that provides reasoning traces. The model to be distilled from, termed the external teacher, first provides a reasoning trace as privileged context, which can be generated in an offline fashion. The student generates rollouts without this context, while a copy of the student conditioned on it, serving as the self-teacher, scores the same rollouts in the student's own token space. BSD uses the resulting token-level guidance to adjust updates on unsuccessful rollouts, with a verifier providing the outcome signal. As the external teacher is queried only to produce text, the same training procedure applies to all teacher scenarios: shared-tokenizer, cross-tokenizer, and closed API teachers. Across these scenarios, BSD improves over the base model and outperforms baselines tailored to each scenario, offering a novel and unified pathway for distillation. Code is available: https://github.com/TUMLegalTech/Bridged-Self-Distillation

HarnessRL: Co-Evolving Policies with a Just-in-Time Training Harness for Open-Ended RL

Xiaoqiang Wang, Ting-En Lin, Yuchuan Wu, Pinyi Zhang, Caixin Kang, Ling Li — Tue, 20 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 92%

Rubric-based reinforcement learning (RL) trains policies for open-ended generation with natural-language criteria as rewards, but these rewards can become ineffective as training progresses. We identify two failure modes. Under non-discriminative rewards, scores collapse to uniformly low or high values due to insufficient exploration or a mismatch between task difficulty and policy capability. Under spurious rewards, higher scores do not correspond to better responses. Both reflect a training configuration whose task sampling, exploration, and evaluation stay fixed as the policy evolves. We introduce HarnessRL, a training-time harness that co-evolves with the policy through three interfaces for task sampling, rollout guidance, and rubric criteria. HarnessRL maintains a training-time memory from which a harness evolver retrieves within-task history and cross-task evidence. Using this evidence, the evolver attributes each reward failure to the interface whose update can address it. It then runs a proposer and just-in-time critic loop in which the critic accepts a candidate update only after measuring its effect on the current frozen policy, returning rejected candidates for revision. We also curate Rubric-Mix-10k, a five-domain training and evaluation suite for open-ended RL across medical, science, writing, role-play, and instruction-following tasks. On Rubric-Mix-10k, HarnessRL outperforms the strongest baselines by 2.8 to 3.1 points across both 4B and 9B model scales and nearly halves the training steps to its best checkpoint. Code and data are available at https://github.com/Tongyi-ConvAI/Qwen-Character/tree/main/HarnessRL.

Mitigating Reward Hacking in Preference Reward Models with Evolving Penalties

Afonso Baldo, Ant\'onio Farinhas, Marcos Treviso, Ricardo Rei, Nuno M. Guerreiro — Tue, 20 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 43%

Reinforcement learning for language models has moved from preference reward models to verifiable rewards, which sustain much longer optimization. Preference models are prone to reward hacking: the policy raises their score without improving response quality, or even while degrading it. Yet for open-ended tasks, where quality can neither be verified nor fully written down as rubrics, they remain valuable. We introduce Dynalize, which mitigates reward hacking online, during policy training. Periodically and asynchronously, a language model analyzes training rollouts and writes penalties for undesirable behaviors that the reward model appears to reward; a judge applies them to every rollout. In clinical counseling and creative writing, training on the reward model alone raises its score continuously but ultimately leads to a decline in external benchmarks. Dynalize sustains longer training and results in improved downstream performance. For example, we find that our method suppresses patterns incentivized by preference models that, without penalties, spread to most responses, and even to the model's reasoning, which the reward model does not explicitly score. Furthermore, we show that its gains hold in multi-turn training, and even after we retrain the reward model on the discovered behaviors.

Revision-Based Learning from Interactive Experience and Feedback

Jiawen Li, Chenming Ge, Alice Gao, Rishab Vempati, Yi-Hao Peng, Jason Wu — Tue, 20 Oct 2026 00:00:00 +0000

93% · New model 93%, Evaluation 79%, Agents 39%

Code generation models are increasingly used to produce interactive artifacts, such as websites and user interfaces. Many properties that determine whether these interfaces are useful only become apparent through realistic interaction, making them difficult to incorporate into model training. In this paper, we introduce RELIEF (REvision-based Learning from Interactive Experience and Feedback), a method for improving code generation models using feedback on their generated interfaces. RELIEF uses a coding agent to revise these interfaces, then incorporates the improvements into their original generation traces to produce supervision for fine-tuning. To evaluate RELIEF, we collected a dataset of 200 synthetically generated websites paired with 355 interaction recordings and accompanying feedback from 24 users. We used critiques derived from code inspection, automated interaction, explicit human feedback, and human interaction recordings to fine-tune variants of two code generation models. Our results show that feedback effectiveness varies with base model capabilities, with critiques inferred from human interaction recordings leading to the largest usability improvements overall under our tested models.

Competing Learning in Hybrid Architectures

Xingyu Shen, Yingfa Chen, Ziqing Qiao, Chaojun Xiao, Xu Han, Zhiyuan Liu — Tue, 20 Oct 2026 00:00:00 +0000

93% · New model 93%

Many modern language models have adopted hybrid architectures that incorporate both linear-time recurrence and softmax attention. Modern recurrence layers provide a highly expressive mixing mechanism while ensuring high throughput by compressing contextual information into a fixed-size state, but underperform softmax attention in contextual information retrieval. In this paper, we reveal a Competing Learning phenomenon where different kinds of mixers compete to learn the same function, and when any layer learns a function, it suppresses the corresponding learning signal for other layers. Layer attribution analysis reveals that under typical data distributions, modern recurrence learns retrieval functions faster than softmax attention, causing attention to fail to learn retrieval in most cases. Stopping weight updates in recurrence in the middle of training helps attention learn retrieval. Based on these results, we propose a staged training procedure for pre-training hybrid architectures on natural language. Empirical results on models up to 1B parameters show that our method results in substantial improvement in language modeling loss, common-sense reasoning, as well as a wide range of retrieval-related tasks across different context lengths.

ReTTT: Reflection-Driven Test-Time Training via Experience-Based Parametric Memory

Kangcong Li, Hangfan Zhang, Shuyue Hu, Shengji Tang, Tao Chen — Tue, 20 Oct 2026 00:00:00 +0000

92% · New model 92%, Agents 69%, Evaluation 40%

LLM agents accumulate valuable task-specific information through their actions, observations, and environmental feedback during multi-turn task execution. When effectively used, this information can help agents refine subsequent decisions and improve task performance. However, the learning value of an interaction depends on its context and subsequent feedback, making it challenging to extract reliable, actionable lessons and revise them as the task unfolds. In this paper, we introduce ReTTT, which enables experience-driven self-adaptation through test-time training on lessons constructed and continually revised within an episode. ReTTT lets the agent reflect on interaction feedback, maintain the resulting lessons in an explicit memory, and learn from them by updating an episode-local LoRA adapter while keeping the backbone frozen. The adapted model then performs both subsequent actions and reflection, coupling policy adaptation with the construction of future learning experience. We evaluate ReTTT on ALFWorld, ScienceWorld, and SWE-bench Lite using two model families. ReTTT outperforms trajectory-based test-time training across all six model--benchmark settings, improving task success by up to 12.14 percentage points. A frozen-reflection comparison further shows a 7.14-point benefit from adapting the reflection policy, while trajectory analysis illustrates how adapted reflection derives more task-specific guidance from failed attempts. These findings suggest that learning to reflect on its own behavior can boost agents' performance on long-horizon tasks.

Compaction-Aware Reinforcement Learning For Ascend Triton Kernel Generation

Chuansai Zhou, Yanlin Liu, Bingning Huang, Shidi Xi, Chao Huang, Boxing Chen — Tue, 20 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 73%, Agents 61%

Automatically generating high-performance kernels with LLM-based agents promises substantial gains in development efficiency, but two obstacles remain. First, while increasing multi-turn interaction with execution environments is crucial for refining generation quality, continuous context accumulation rapidly exceeds the model's effective context window. Second, training long-horizon tasks via GRPO inherently incurs sparse rewards due to its reliance on terminal outcome evaluation, severely hindering step-level credit assignment; To address these obstacles, we propose ATRL (Ascend Triton Reinforcement Learning), a post-training method that refines GRPO into a compaction-aware, segment-level objective for long-horizon kernel generation. We also design a lightweight agent harness providing cascaded compilation, correctness, and profiling feedback, alongside an Ascend-specific knowledge base injected during agentic RL training and evaluation to compensate for the base model's missing platform expertise. Evaluations on NPUKernelBench and KernelBench, both executed on Ascend hardware, show that ATRL consistently surpasses all baselines, including GRPO, in both correctness and speedup.

Preserving Reasoning Diversity in Reinforcement Learning with Verifiable Rewards via Adaptive Replay

Wen-Zhi Li, Sainyam Galhotra, Claire Cardie — Tue, 20 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 36%

Reinforcement Learning with Verifiable Rewards (RLVR) improves the reasoning capabilities of large language models, but often reduces the diversity of their reasoning behaviors. We study this phenomenon from the perspective of finite-sample learning dynamics and show that limited on-policy rollouts can progressively suppress semantically distinct but equally rewarded reasoning modes. Once a mode becomes sufficiently under-sampled, standard on-policy RLVR rarely recovers it. Motivated by this observation, we propose TAR, a Targeted Adaptive Replay framework that preserves diverse correct reasoning behaviors already present in the base policy. TAR constructs a compact memory bank of diverse base-model solutions and selectively replays them through an adaptive mechanism that activates only when stored behaviors are being suppressed during RLVR. Across six mathematical reasoning benchmarks and two Qwen2.5 model scales, TAR consistently achieves a better balance between reasoning performance and diversity preservation than existing diversity-preserving post-training methods, while substantially reducing the forgetting of correct base-model solutions. Code is available at https://github.com/wenzhilics/tar.

ProbePO: Sparse Probes, Redistributed Credit for Critic-Free LLM Reinforcement Learning

Diyi Hu, Ruizhong Qiu, Lin Qiu, Hanghang Tong, Hanqing Zeng — Tue, 20 Oct 2026 00:00:00 +0000

92% · New model 92%, Efficiency 69%, Evaluation 57%

Reinforcement learning from verifiable rewards assigns one terminal reward to an entire reasoning trajectory, obscuring how individual reasoning steps contributed to the outcome. GRPO broadcasts a group-normalized advantage to every response token, while finer-grained approaches often rely on dense auxiliary rollouts or branching schemes that can also alter the data used for policy updates. We cast the missing middle as budgeted measurement: with a limited rollout budget, which partial solutions should be evaluated, and how should their value estimates refine credit? ProbePO selects partial-solution states with a reward-free uncertainty score and estimates their expected verifier rewards under a capped continuation budget. ProbePO redistributes, rather than replaces, GRPO's trajectory-level advantage across tokens while preserving each trajectory's mean token-level advantage; probe continuations serve only as measurements. Across three DeepSeekMath and Qwen-family models spanning 1.5B-7B parameters, ProbePO consistently outperforms GRPO and achieves the highest accuracy among the four evaluated methods in 16 of 18 model-benchmark comparisons. Against dense-probing VinePPO on DeepSeekMath-7B, ProbePO achieves higher accuracy on all six benchmarks with 76% less end-to-end training time.

Verifying Human-Written Reasoning References

Ali Ansari, Farhana Ferdousi Liza — Tue, 20 Oct 2026 00:00:00 +0000

92% · Evaluation 92%

Evaluating the reasoning of AI systems often relies on human-written reasoning chains as references, yet the logical support for their individual steps is rarely verified. Although prior neuro-symbolic approaches have formalized and checked reasoning, their reported evaluations do not systematically assess whether step conclusions in human-written reference derivations are entailed by their cited premise sets or investigate the sources of failures in those references. To address these limitations, we present a novel reference-centred assessment method for human-written derivations that combines entailment checking against cited premise sets, full-premise rechecking, and failure attribution to examine the reliability of these derivations and their formal representations. We evaluate it on P-FOLIO, using a large language model (LLM) to translate step conclusions into first-order logic (FOL) and the Z3 SMT solver to check entailment. Of 386 derivations examined, 202 (52.3%) meet all our checking criteria, meaning that every translated step is entailed by its stated support and the required target is established under the available FOL representations. Among 230 steps not entailed by their stated support, 191 become entailed when checked against the complete original premises, indicating that most initial failures reflect incomplete or mismatched support rather than unsupported claims. Manual analysis of the remaining 39 cases identifies 21 human reasoning errors, 11 FOL translation errors, and 7 dataset annotation errors. A separate manual audit of 50 randomly sampled steps that passed the entailment checks found 46 faithful FOL translations (92%). These findings suggest that disagreement with a human-written reasoning reference may originate in the reference or its formal representation rather than in the AI system being evaluated.

Beyond the Teacher: Understanding and Preserving Distillation Overshoot in Agentic On-Policy Distillation

Jike Zhong, Ming Li, Yuxiang Lai, Xi Xiao, Sheng Zhang, Ritwick Chaudhry — Tue, 20 Oct 2026 00:00:00 +0000

92% · New model 92%, Agents 61%

On-policy distillation (OPD) trains agents on their own trajectories using supervision from a stronger teacher. We uncover a counterintuitive phenomenon in multi-turn agentic tasks: a smaller, initially weaker student can substantially outperform its teacher during training, only to regress as distillation continues. We term this phenomenon distillation overshoot. Through decision- and task-level analyses, we trace overshoot to two competing effects of OPD: repairing weak student decisions and eroding successful student choices that the teacher disfavors. Early in training, repair outpaces erosion, allowing the student to combine the teacher's corrections with its own strengths and surpass the teacher. As training continues, erosion overtakes repair, causing these gains to vanish. Motivated by this finding, we propose Success-Preserving On-Policy Distillation (SPOD), which selectively masks conflicting teacher supervision on successful student choices and introduces a soft probability floor to counteract their erosion, while retaining teacher guidance elsewhere. SPOD requires only binary episode success and model probabilities already available during OPD. Across two model families and three agentic benchmarks, SPOD consistently outperforms ordinary OPD and evaluated variants, improving final success by up to 15.2 percentage points and surpassing the teacher by up to 20.0 points. Moreover, SPOD's final checkpoints outperform even ordinary OPD's best intermediate checkpoints in all six settings, demonstrating that beyond-teacher gains can be not only preserved but further extended. Code: https://github.com/Jike338/SPOD

Atoms, Not Knowledge: What Production-Scale Addressed-Memory Writes Deposit

Wei Gao, Cheng Tian, Duan Yang, Jianqing Yin — Tue, 20 Oct 2026 00:00:00 +0000

92% · Evaluation 92%

Deterministic hash-addressed conditional memory permits local post-hoc knowledge injection. Extending arXiv:2610.21553's bounded-feasibility audit rejecting under-specified generality claims with repaired probes and training-dose/capacity axes, we give, to our knowledge, the first production-scale write-content verdict. We test 124M-7B self-trained/publicly pretrained beds, Qwen3.8-Flash-Next-FP8 (176.95B; 320,001,536 FP8 rows), and DeepSeek-V4.1-Flash. Write coverage requires a sufficiently trained backbone prior, full-width readout, trainable gates; undertrained-bed intervention confirms a threshold, not a slope. At fixed budget, coverage falls from 98% (50 facts) to 12.7% (1,000); table size is not binding. Two-hop composition does not exceed the independent single-hop baseline (26% observed vs 25% expected; 15% lexical-distortion floor). Synergy remains absent across dose/load in two independent bed families. A 28-percentage-point in-context gap (given vs recalled intermediates) localizes failure to retrieval-for-reasoning, not gate access. Writes are independently retrievable, precisely addressable, reversible atoms, not composable into reasoning. A layered protocol, the first, to our knowledge, to cover compositionality in behavioral memory evaluation, separates baseline, synergy and distortion floor. Any addressed/key-value memory must report coverage, synergy ratio and distortion rate before deployment claims. Composition must be system-orchestrated; multi-constraint queries should explicitly chain written atoms. In-context controls show given intermediate entities are exploited at twice the implicit rate. Production load fits exponential interference ($R^2=0.98$); its mechanism, between budget dilution and write interference, remains open. All numbers are recomputed from archived raw logs/JSON; a fresh 52-target production rerun gives 52/52, exactly matching archived baselines.

RARE: Reinforcement Learning with Adaptive Rubric Evolution for Open-Ended Generation

Xueyun Tian, Bingbing Xu, Chuyue Li, Jiaming Liu, Chao Zhang, Weijia Feng — Tue, 20 Oct 2026 00:00:00 +0000

92% · New model 92%, Evaluation 62%

Verifiable rewards have advanced reinforcement learning in mathematics and code. However, many valuable open-ended tasks share three properties: no automatic verifier of overall quality, multiple requirements to satisfy jointly, and multiple valid responses to each query. Training therefore needs feedback on requirement satisfaction across different valid responses. Rubric-based reinforcement learning provides this feedback by aggregating scores from explicit criteria. As learning progresses unevenly, some criteria become fully satisfied and lose discrimination while others remain informative. Dynamic methods revise shared rubrics or select evolving criteria to restore discrimination, but may also change useful criteria. To address this, we propose Reinforcement Learning with Adaptive Rubric Evolution (RARE). The core idea is to use current policy feedback to guide both which criteria change and what their replacements assess. Sample-level updates replace criteria fully satisfied across current outputs with checks of remaining quality differences, retaining all other criteria and scores. Within this rule, we compare rewarding desirable qualities with penalizing observed flaws. Video script generation exemplifies these properties and supports content production. Its demands for coherent, executable narratives under constraints make it a challenging testbed. Targeting flaws performs best across three video types with a mean gain of 3.82 points over matched static rubrics. Model and human preferences support these gains. Improvements also extend to writing and general generation benchmarks.

StopLoss: Teaching Agents to Predict Failure and Stop Early

Jiarong Liang, Xueguang Ma, Dongfu Jiang, Yuxuan Zhang, Ping Nie, Wenhu Chen — Tue, 20 Oct 2026 00:00:00 +0000

92% · Agents 92%, Efficiency 91%, New model 70%

Coding agents keep working on runs that will fail verification, and asking a frontier model to watch for this is expensive in its own right. We introduce StopLoss, a small external monitor that learns failure recognition from completed trajectories and their verifier outcomes. The name is inspired by the stop-loss order in finance: StopLoss ends a run early when the run is likely to fail, which saves resources. Final outcomes label earlier local windows, so a fine-tuned language model can score the most recent tool steps in a single forward pass, and a threshold calibrated on successful runs turns that score into an abort decision without changing the executor. Trained on GPT-family SWE-rebench trajectories and evaluated on SWE-bench Verified and a SWE-bench Pro subset with GPT and Claude executors, StopLoss is compared against six general-purpose judges reading identical windows. Outcome supervision lifts failure prediction from near chance to a useful signal at the first check, at one to two orders of magnitude lower cost per check than any tested judge. This translates into savings: relative to completing every task, StopLoss lowers cost per retained solve by 4.5\% on Verified while preserving 99\% of successful runs, and by up to 31\% on the costlier Pro subset, reaching 48\% with a Qwen3-8B backbone, and in all four settings its cost per retained solve is below that of every judge policy. Outcome-trained small monitors can make failure recognition cheap enough for early stopping to pay for itself.