Paper Radar · live feeds

Paper Radar · today in AI

Today's standouts across cs.AI, cs.CL and cs.LG, picked by Jev from every new submission.

262 read → 50 shortlisted → 32 worth opening · 2026-11-09 · $0.0103

· AlienBench: A Benchmark of Structural Novelty in Language Model Outputs
· FondueQA: A Benchmark for Evaluating Parametric Compensation in Multi-Hop QA
· One Run Is Not Enough: A Monte Carlo Protocol for Evaluating LLM Reliability on Data Science Tasks

Tool-using and web-browsing LLM agents, from every new cs.AI, cs.CL, cs.LG and cs.MA submission.

267 read → 11 shortlisted → 4 worth opening · 2026-11-09 · $0.0102

· Debate to Deceive: Poisoning LLM's Fact-Checking via Game-Theoretic Multi-Agent Debate
· Rethinking Agent Trajectory Compression: A Workflow-Aware Perspective
· AIM: Adaptive Interaction in Multi-Agent Debate for Multimodal LLM Inference

Machine learning evaluated on patients, from new PubMed records — not preprints.

185 read → 17 shortlisted → 13 worth opening · 2026-11-09 · $0.0085

· Development and validation of a habitat-based radiomics fusion model for predicting immunotherapy response in advanced NSCLC: a multicenter retrospective study.
· Prediction of incident inflammatory bowel disease from routine electronic health records: development and external validation of a longitudinal machine learning model.
· 18F-FDG PET/CT Habitat Radiomics for Predicting Epidermal Growth Factor Receptor Mutation Status in Locally Advanced Non-small Cell Lung Cancer: A Multicenter Study.

Making inference faster, smaller or cheaper — across cs.LG, cs.DC and cs.AR.

146 read → 9 shortlisted → 4 worth opening · 2026-11-09 · $0.0056

· Knowledge Distillation from Neural Networks to Differentiable Logic Networks
· Nebula: Distributed Position-Independent Caching for LLM Serving
· DARIC: A Position-Independent KV Cache Reuse System with Distribution-Aware Recomputation

New bioRxiv neuroscience preprints, filtered for computational and systems work.

10 read → 0 shortlisted → 0 worth opening · 2026-11-09 · $0.0004

Learned control and manipulation, from every new cs.RO submission.

54 read → 2 shortlisted → 1 worth opening · 2026-11-09 · $0.0021

· Grasp-TAGS: Task-Aware Grasp Selection via Critic Distillation for Prehensile Loco-Manipulation

Tech news with the retellings removed: first-party announcements, measured numbers, and things you can use today.

259 read → 25 shortlisted → 15 worth opening · 2026-11-09 · $0.0098

· Joined the 4x R9700 Club | Initial Benchmarks for QFN 3.8 Below
· Motion graphics video made with Qwen3.8 Flash Next GSQ RCO IQ3_XXS on strata
· MIT CSAIL: GPT-6 Astra edited 0.016% of a small Qwen model's weights and lifted MATH-500 from 7.9% to 48.1% -- critics say it's standard ML engineering