Jianbin Zhang, Xin Sun, Shanwen Wang et al. · cs.LG
Compression 98%
Abstract
Knowledge distillation is crucial for deploying vision-language models on resource-constrained devices. However, existing methods typically impose uniform supervision across visual tokens or rely on static token selection, which confuses task-relevant cues with background noise and degrades cross-modal reasoning. To address this limitation, we propose Key Visual Evidence-guided Knowledge Distillation (KVE-KD), a framework that dynamically focuses feature distillation on task-relevant visual tokens identified by the teacher model. Specifically, KVE-KD appoints the final pre-generation textual token as a unified semantic anchor and identifies the target cross-modal fusion layer by analyzing changes in the anchor representation through iterative visual-token contribution removal. Within this layer, KVE-KD ranks visual tokens via the anchor-conditioned attention distribution and selects the most informative visual tokens as key visual evidence with normalized entropy. The key visual evidence subsequently guides focused visual feature distillation, making the student align closely with the teacher's task-relevant visual representations while suppressing irrelevant background information. Extensive experiments on six benchmarks demonstrate that KVE-KD outperforms state-of-the-art cross-modal distillation methods, with particularly pronounced gains on tasks requiring complex reasoning and fine-grained visual understanding. Importantly, these improvements are achieved without introducing any inference-time overhead. The source code is available at https://github.com/zhangjianbin07/KVE-KD.
Adil Mubashir Chaudhry, Osama Ahmad, Zubair Khalid et al. · cs.LG
Compression 96%Inference speed 30%
Abstract
Model compression strategies are widely employed to reduce memory footprint and network complexity, particularly for devices with constrained computational, memory, and energy resources. Prior works that rely on simultaneous conversion from floating-point high-precision (FP32) to integer low-precision (INT4) representations and distillation into smaller models suffer from unstable training and drastic degradation of prediction performance. To address these limitations, we propose a unified framework, known as \textbf{P}rogressive \textbf{M}ulti-\textbf{A}ncestor \textbf{B}it-depth \textbf{D}istillation (PMABD), that progressively compresses the network while transferring knowledge through a growing pool of higher-precision ancestor teachers. PMABD generates a sequence of intermediate teachers that each learn from all higher-precision ancestors and jointly supervise the final target student. This multi-ancestor, multi-stage design stabilizes ultra-low-bit quantization by lowering quantization noise profiles across training and ensuring stable quantization. Experiments on CIFAR-10/100 with ResNet-20/32/18, and Tiny-ImageNet with MobileNetV2 show that PMABD outperforms state-of-the-art compression frameworks, results in 1.06$\%$ increase in performance of W2A2 (ResNet-18/CIFAR-100) student model. We show that a saturation-based stopping criterion contributes to improve the performance of our final student.
As large language models (LLMs) grow in parameter count, model storage and parameter memory traffic have become major bottlenecks to efficient deployment. Codebook-based weight quantization reduces these costs, but imbalanced nearest-codeword assignments during fitting can leave some codewords insufficiently updated, limiting effective codebook utilization. To address this limitation, we propose Balanced Assignment Refinement for Quantization (BARQ), which improves quantization quality through balanced fitting of existing codebooks. Specifically, we first compute joint soft assignments between weight blocks and codewords through entropically regularized optimal transport with uniform marginals and curvature-weighted reconstruction costs, ensuring equal positive fitting mass for every codeword in the exact solution. We then refine the codewords through an assignment-weighted barycentric update, which we prove minimizes the fitting objective for fixed assignments. For finite Sinkhorn iterations, the implemented update retains this optimality provided all codeword masses exceed the denominator floor. Finally, we discard the soft assignments and use the refined codebook for standard hard nearest-codeword encoding, with our analysis establishing sufficient conditions for reducing hard-quantization distortion and evaluation loss. Across multiple LLMs, BARQ achieves lower perplexity and higher mean zero-shot accuracy than the evaluated baselines at comparable bit budgets. The code is available at https://github.com/chenhangcuisg-code/BARQ.
Weiling Yang, Junwen Zhang, Dezun Dong et al. · cs.DC
Inference speed 95%Serving systems 73%
Abstract
Mixture-of-Experts (MoE) inference is increasingly deployed in local and on-premise environments, where expert parameters often exceed GPU memory capacity. In latency-sensitive, low-concurrency settings, repeatedly staging routed-expert weights from CPU memory to the GPU can be prohibitive, leaving routed-expert feed-forward networks (FFNs) on the critical path of multi-socket CPUs. Existing CPU accelerations often rely on intrusive, hardware- or topology-specific requirements, such as AMX-specific weight layouts or manual NUMA-aware placement. These requirements reduce portability and complicate integration with standard CPU-GPU offloading pipelines. We present HiNa-MoE, a high-performance, non-intrusive operator library for MoE inference on CPUs with Intel AMX. HiNa-MoE (1) exploits AMX with an optimized micro-kernel that keeps expert weights in standard layouts and instead fuses lightweight layout transforms into token gathering and stores; (2) applies NUMA-aware task partitioning under a simple page-interleaved policy without modifying the framework allocator; and (3) converts decode-phase memory matrix-vector operations into small matrix-matrix execution to utilize AMX. Across multiple MoE models, HiNa-MoE achieves up to 3.37x speedup for FFN kernels and up to 2.09x end-to-end inference speedup over state-of-the-art baselines, while remaining plug-and-play with existing frameworks and deployment workflows.
Looped language models provide a parameter-efficient way to scale iterative test-time computation by repeatedly executing a shared recurrent core. Post-training quantization (PTQ) can reduce the memory footprint and inference cost of looped language models, but errors introduced by a quantized shared core affect subsequent cores. Among PTQ methods, channel scaling and orthogonal rotations preserve the floating-point computation while producing representations with different quantization quality. We find that quantization configuration candidate rankings can change with recurrent depth, motivating configuration selection at the target deployment depth. However, evaluating every candidate over the full calibration set at this depth is costly. We therefore propose Loopy, a PTQ framework that formulates shared-core quantization through a recurrent-depth-aware objective, selecting shared low-bit representations by their final prediction loss at the target deployment depth. Channel scaling and orthogonal rotations parameterize the candidate representations. To approximately solve this selection problem efficiently, Loopy progressively allocates calibration windows to promising candidates while preserving complete target-depth execution, using only forward evaluations. Across eight settings, Loopy achieves the state-of-the-art results among different baselines. On Ouro-1.4B under W4A4, Loopy reduces LAMBADA perplexity by 36.5% relative to SpinQuant. Our code is available at https://github.com/Shameless0817/Loopy-review.git.
Block drafters such as DSpark propose ranked candidates for several positions in one forward pass, and a tree verifier checks them in one pass of the target. The number of rows to verify trades the tokens a wider tree is expected to accept against the time a wider verify takes. Most schedulers that choose this number take the verify time from a table or model measured before serving, corrected online by at most one scale factor, and take acceptance from the drafter's confidence estimates or from a map fitted offline. AdaSpark learns both quantities while it serves, with no profile, calibration or sweep in advance. It learns which verify widths are worth offering and fits each one's verify time as a function of context. It fits each candidate's acceptance probability to the target's verify outcomes, with the drafter's confidence head as one input, and orders and sizes the tree by that fit instead of by the head. The same model prices n-gram continuations of the request's own text, so drafted and text-derived candidates compete for rows in one best-first order. The width is chosen by pricing time at the long-run decode rate. On single- and multi-turn conversations from six public datasets, on three dense targets and one mixture-of-experts target, AdaSpark decodes 1.5-3.1x faster than llama.cpp's DSpark with the same drafters. Our imparo engine with AdaSpark is 1.17-1.52x faster than imparo running with a three-token chain (the default llama.cpp setting); this gain comes from the scheduler alone. Without a width sweep, AdaSpark is never more than 0.3% slower than the best pinned tree width on any dense target or context band. On the mixture-of-experts target it ties the best pinned width, and the other pinned widths from 4 to 16 rows are 5-14% slower.
Serving a large language model (LLM) across a fleet of deployments requires several weight-precision operating points. Multi-precision formats serve them all from one stream whose prefixes are valid lower-precision codes, instead of storing multiple copies. We present StagQ, a multi-precision weight format whose main stream is a 2-bit group-wise affine base followed by a configurable number of 1-bit refinement planes on a dyadic step schedule. Every supported precision is a readable prefix, decoded by an affine map derived from metadata shared across all precisions, with no per-weight lookup. A sparse side record, filled both before and after the grid is fitted, holds out the few weights the grid serves worst. We report two configurations of the encoder. At two bits the cheaper one leads the strongest multi-precision baseline on Llama-3.1-8B, Phi-4, and OLMo-2-7B by 3.1 to 7.0 MMLU points, at a slightly lower logical rate. At three bits it leads on Llama-3.1-8B, leads on Phi-4 at a higher rate, and ties on OLMo-2-7B. At four bits it ties on all three, at a higher rate. In a batch-one matrix-vector product on an NVIDIA A100 GPU, timed on synthetic weights, our kernel is faster than the two baseline kernels in most shape-precision cases.
A language model can give a correct answer more probability than any single incorrect answer and still usually sample an incorrect one, because the incorrect answers together hold more probability. The power distribution raises each complete answer's probability to a power above one and renormalizes, shifting probability toward answers the model finds most likely (sharpening). Sampling from it improves reasoning without changing parameters, but needs many scored candidates per query. We show that a model can instead be trained to produce such answers in one generation. On-policy power distillation (OPPD) runs a sequential Monte Carlo sampler in which the model being trained generates candidates and a frozen teacher's power distribution weights them; the same probabilities weight each answer in a maximum-likelihood update. Training raises single-generation accuracy by up to 23.0 points on MATH500 and 27.3 on GSM8K over the untrained model at the same temperature, and one generation scores 2.4 and 3.5 points above published power sampling with 64 candidates, recovering 94 percent of the gain that 16 candidates give the untrained model. For context, against GRPO trained with verified rewards from the same checkpoint and budget, OPPD scores 3.8, 4.0 and 5.4 points higher on MATH500, GSM8K and AIME using no reference answers; the two are complementary, and OPPD applied after GRPO adds up to 9.3 points. Trained only on mathematics, OPPD raises HumanEval accuracy by up to 5.3 points. One loss coefficient moves the sharpening exponent the model absorbs between 1.19 and 2.02, against 1.14 for ordinary on-policy distillation, and it rises mostly on the model's own answers. Gains hold across model families and sizes, including a model already trained with verified rewards, where lowering the temperature gives nothing and OPPD adds 4.4 points on MATH500. Code: https://github.com/ArminAzizi98/OPPD.
On-policy self-distillation (OPSD) provides dense token-level supervision without a second model: one network acts as teacher with the reference solution and as student with only the problem. We identify a specific failure mode of this recipe. During training, student token entropy rises past the teacher's and remains elevated, a pattern we call entropy overshoot. We trace it to both sides of distillation. The reference-conditioned teacher is confident along its answer-directed reasoning path, but this confidence transfers poorly to student-generated prefixes, making its supervision overly tied to answer-specific cues rather than reusable reasoning patterns; meanwhile, the forward KL used by OPSD continually diffuses the student's predictive distribution without pulling it back. We introduce E$^2$-OPSD to address both causes. Exemplar-guided teaching replaces the current answer with a retrieved solved neighboring problem, providing transferable reasoning guidance without revealing the destination and better matching student-reachable states. Entropy-aware distillation uses the student-teacher entropy gap to determine the direction and strength of each token's correction. E$^2$-OPSD improves math reasoning by up to 4.3 points in mean@16 over OPSD, while out-of-domain evaluations show gains over the corresponding base models of up to 4.9 points in mean@16 and 5.5 points in pass@8. Despite these gains, E$^2$-OPSD remains simple, requiring no additional forward passes or networks.
ZheXu Wang, Mao-Lin Luo, Yankun Hong et al. · cs.LG
Compression 94%
Abstract
On-policy self-distillation (OPSD) provides denser token-level supervision and better computational efficiency than Reinforcement Learning with Verifiable Rewards (RLVR). However, this denser supervision may introduce substantial noise and training instability. Existing improvements often rely on high-variance per-token statistics and introduce extra hyperparameters and trade-offs. Based on the advantage formulation in RLVR, we analyze the OPSD objective from the same perspective, incorporating outcome correctness signals. We find that vanilla OPSD imposes insufficient penalties and excessive rewards on incorrect trajectories because it applies a fixed divergence objective regardless of outcome correctness. Furthermore, the reliability of teacher supervision is associated with both trajectory outcome and the cumulative average teacher entropy along the rollout. Based on these observations, we propose Outcome-Guided On-Policy Self-Distillation (OG-OPSD), which dynamically adapts both the divergence objective and distillation position according to binary outcome rewards and the cumulative average teacher entropy. Extensive experiments show that OG-OPSD consistently improves the performance of vanilla OPSD and multiple strong baselines in mathematical reasoning, multimodal reasoning, and out-of-distribution tasks across Qwen3 models at 1.7B, 4B, and 8B scales, as well as Qwen3-VL-2B.
Penghui Yang, Long Xing, Xuanlang Dai et al. · cs.LG, cs.AI
Compression 93%
Abstract
On-policy distillation (OPD) trains a student model on self-generated trajectories, but transmitting dense teacher distributions across long reasoning traces creates prohibitive communication and memory bottlenecks. Practical systems therefore rely on sparse teacher interfaces, typically transmitting either the sampled-token score or a small Top-$k$ distribution. However, this sparse setting faces a fundamental dilemma: sampled-token estimators are unbiased but suffer from severe gradient variance, whereas directly optimizing Top-$k$ objectives introduces systematic bias. To improve this trade-off, we propose ResOPD (On-Policy Distillation with Tail Residualization), which provides unbiased full-vocabulary reverse KL gradient estimation under on-policy sampling, with substantial variance reduction under the same sparse payload in the evaluated settings. ResOPD aggregates the unobserved vocabulary into an observable coarse tail event, computes its exact aggregate gradient, and samples only the fine-grained within-tail residual, which requires no additional teacher queries or forward passes. Extensive experiments demonstrate that ResOPD substantially reduces gradient variance, stabilizes online training dynamics, and improves downstream performance across the evaluated settings. These results establish ResOPD as an efficient, plug-and-play variance reduction primitive for sparse on-policy distillation.
Jafar Badour, Maurice van Keulen, Elena Mocanu · cs.LG, cs.DC
Inference speed 91%Serving systems 67%
Abstract
Deep neural networks continue to grow in parameter count, driving up training and inference cost on GPUs. Sparse neural networks and Dynamic Sparse Training (DST) promise to reduce these costs, but most implementations rely on binary masks over dense tensors and recover little of the theoretical compute, memory, or energy savings. We propose SNACK, a truly sparse GPU layer that stores and computes only non-zero connections. SNACK exposes a simple PyTorch API for restructuring connections and backpropagating gradients entirely in the sparse paradigm, and ships SNACK-COO, a custom COO-format SpMM CUDA kernel with a batch-to-Streaming-Multiprocessor mapping tuned for the small-batch, high-sparsity regime typical of large-model training and single-stream inference. At the kernel level, SNACK is up to 7x faster than the masked dense baseline (Dense+Mask) and competitive with cuSPARSE, Sputnik, and FlashSparse at 95% sparsity. At 90% sparsity, a single SNACK layer accelerates training by 8x and 3.7x, and inference by 4x and 2x, over Dense+Mask and fully dense layers, respectively, while using 72% less memory than dense and substantially less energy. End-to-end, SNACK reduces GPT-2 peak training memory by up to 40% and graph-style inference latency by 4.8x over Dense+Mask at 99% sparsity.
Karn Tiwari, Varnith Chordia, Prathosh A P · cs.LG
Compression 91%
Abstract
On-policy distillation (OPD) has emerged as a widely used paradigm for post-training large language models, reducing the train--test mismatch of conventional distillation by supervising the student on its own generated trajectories. However, existing OPD objectives remain largely token-local and outcome-agnostic, optimizing teacher--student agreement at each prefix despite reasoning quality being determined at the trajectory level. Reinforcement learning with verifiable rewards (RLVR), particularly Group Relative Policy Optimization (GRPO), provides complementary outcome-level supervision but suffers from sparse rewards and coarse credit assignment. We show that OPD and RLVR exhibit complementary blind spots: teacher signals provide dense local guidance but are weakly aligned with rollout correctness, whereas group-relative rewards capture task success but provide coarse token-level credit and vanish on all-failure groups. We introduce DiffGate, an outcome-gated objective that combines GRPO with selective, bounded teacher guidance. Teacher supervision is applied only to failed trajectories, scaled by group difficulty, and smoothly bounded to prevent extreme teacher--student discrepancies from dominating optimization. The verifier therefore determines \emph{which trajectories} receive teacher guidance, while the teacher provides dense token-level update directions within those trajectories. Across Qwen3-0.6B and Qwen3-1.7B students, DiffGate improves code avg@8 over matched GRPO by $+1.7$ and $+1.8$ points and pass@8 by $+1.6$ and $+5.7$ points, respectively. On mathematics, avg@8 remains within $0.5$ points of GRPO while pass@8 improves by $+1.1$ and $+3.9$ points. Overall, DiffGate improves pass@8 across all four model--domain settings, demonstrating improved solution coverage under our evaluation protocol.
Shen Ruan, Wenchang Gao, Jin Wang et al. · cs.LG, cs.RO
Compression 91%
Abstract
Post-training quantization has been shown to preserve VLA performance under standard evaluation conditions, but whether it preserves the full-precision model's robustness and generalization remains underexplored. In this study, we systematically study the robustness and generalization of post-quantized VLA policies under environmental disturbances. Empirical results show that quantized policies can become fragile to subtle environmental variations despite retaining comparable in-distribution performance. We further observe that action discrepancies are concentrated in a small subset of rollout states, while teacher guidance has opposite effects depending on discrepancy: it improves generalization at high-discrepancy states but can degrade it at low-discrepancy states. These findings reveal that effective post-quantization recovery requires selectively intervening on vulnerable states rather than globally distilling the student. We therefore propose Policy-Induced Vulnerability-Oriented Tuning (PIVOT-Q), a vulnerability-aware On-Policy Distillation (OPD) framework that selectively corrects vulnerable states encountered during quantized-student rollouts using the frozen full-precision policy as a teacher. PIVOT-Q identifies vulnerable states using discounted accumulated discrepancies over a short horizon, applies phase-balanced sparse supervision, and uses a Behavioral Anchor to prevent unnecessary changes. Experiments under seven LIBERO-Plus environmental variations demonstrate consistent recovery across multiple VLA backbones and quantization methods. Notably, PIVOT-Q consistently outperforms full-state distillation across all settings while using only 7.4% of its state-level distillation budget. Our code is available at https://github.com/ruanruan-andy/PIVOT-Q.
Juan Garcia Giraldo, Matteo Santelmo, Eduard Durech et al. · cs.LG
Compression 91%
Abstract
Small language models are often post-trained as students on reasoning traces from stronger teacher models to efficiently learn new skills. However, token-level imitation on traces that lie far outside the student's expected distribution often produces \textit{confident conflicts}, whereby the student is required to imitate a continuation that it deems unlikely (i.e., low-probability) despite being confident in a different continuation (i.e., in a low-entropy state). To mitigate the degradation in generalisation and catastrophic forgetting caused by these conflicts, we propose \textbf{Entropy-Aware Mixing}: a dynamic per-token interpolation of the student and teacher distributions, gated by the student's predictive entropy. We implement both convex and geometric interpolations for both offline trace generation (via speculative decoding, then SFT) and on-policy forward-KL distillation. Our results show that entropy-aware mixing stabilises distillation, improving in-distribution and out-of-distribution math reasoning while better preserving general capabilities than fixed-teacher supervision. Nonetheless, the optimal entropy schedule depends on the training source, with offline-generated traces favouring concave schedules (greater overall teacher influence) and on-policy training favouring linear or convex schedules (teacher concentrated in high-entropy states).
Xiaoyu Chen, Bo Shao, Tiangang Zhu et al. · cs.LG, cs.AI
Compression 90%
Abstract
Reinforcement learning can substantially improve a reasoning teacher, but it is unclear which of those improvements survive when the teacher supervises a smaller on-policy student. We study this question in mathematical reasoning by comparing teacher lineages before and after GRPO, multiple student scales, direct GRPO, and several on-policy distillation objectives. The central finding is that transfer is structured rather than scalar: teacher strength alone does not make dense distillation competitive, while an RL-improved teacher creates useful but metric-dependent student gains. This motivates Transfer-Stratified On-Policy Distillation (TS-OPD), which screens training problems by the joint sampled success of the student and teacher, routes acquisition problems to gated forward KL, routes consolidation problems to gated reverse KL, and adds an entropy brake to protect sampled coverage. Across the main comparison, TS-OPD is the strongest student objective for macro average correctness with the GRPO-improved teacher, while pass@K remains more mixed. Ablations show that the gains come from routing and token gating rather than skipping problems. These results support a transfer-aware view of OPD: stronger teachers help when the supervision direction and token budget match the student's observed ability, not merely because the teacher endpoint is stronger.
On-policy distillation (OPD) provides dense teacher supervision on student-generated trajectories, but generating and evaluating long rollouts incurs substantial training cost. Existing acceleration methods reduce this cost through open-loop rollout schedules or closed-loop horizon adaptation. However, supervision compatibility can vary substantially across trajectories, making a single rollout horizon difficult to match their heterogeneous reliable lengths: an overly short horizon may truncate useful supervision, while an overly long one wastes computation beyond reliable regions. Our key insight is that the trajectory-specific reliability boundary need not be predicted before generation. By viewing reliability as the first-passage of accumulated low teacher--student compatibility events, the boundary is inherently unknown before sampling, yet whether it has been reached can be determined exactly from the observed prefix. Building on this insight, we propose *Flash-OPD*, which shifts from rollout-horizon control to adaptive trajectory-level boundary verification. *Flash-OPD* interleaves cached generation with teacher verification and independently stops each trajectory according to its observed compatibility events. To reduce verification overhead, the recent event rate is used only to schedule the next verification point, while the actual stopping decision always relies on the exact cumulative count. This separation prevents estimation errors from causing premature termination while enabling efficient verification during generation. Extensive experiments across diverse datasets and teacher--student settings show that *Flash-OPD* achieves $2.2\times$--$7.5\times$ speedups over standard OPD while maintaining or improving accuracy.
Long context inference with large language models becomes increasingly expensive as attention must operate over an ever growing KV cache. Page sparse attention reduces this cost by representing each KV page compactly and retrieving only a subset for each query. Existing retrieval methods are designed to estimate attention scores or page relevance, but their objectives do not directly account for how approximation errors affect the resulting value weighted attention output. We introduce \method{}, an output aware page encoding derived from the joint structure of keys and values while preserving the key information needed for accurate retrieval. \method{} is training free and requires no additional value dependent statistics at inference time. Once constructed, its stored representation has the same size and decode time scoring cost as a key only spectral representation. Across long reasoning, long context understanding, and long generation benchmarks, \method{} consistently improves over the key only spectral baseline and performs competitively with recent KV cache compression and retrieval methods. On long reasoning benchmarks, it achieves strong avg@\(k\) performance across model benchmark pairs, while matching or surpassing leading baselines on several long context understanding and generation settings with modest decoding overhead. Code is available at \url{https://github.com/Ashkan13776/oval-kv}.