paper-with-me

Papers

Unifying Group-Relative and Self-Distillation Policy Optimization via Sample Routing

2026-04-02 · Gengsheng Li, Tianyu Yang, Junfeng Fang, Mingyang Song, Mao Zheng, Haiyun Guo, Dan Zhang, Jinqiao Wang, Tat-Seng Chua arxiv

Reinforcement learning with verifiable rewards (RLVR) has become a standard paradigm for post-training large language models. While Group Relative Policy Optimization (GRPO) is widely adopted, its coarse credit assignment uniformly penalizes failed rollouts, lacking the token-level focus needed to efficiently address specific deviations. Self-Distillation Policy Optimization (SDPO) addresses this by providing denser, more targeted logit-level supervision that facilitates rapid early improvement, yet it frequently collapses during prolonged training. We trace this late-stage instability to two intrinsic flaws: self-distillation on already-correct samples introduces optimization ambiguity, and the self-teacher's signal reliability progressively degrades. To resolve these issues, we propose Sample-Routed Policy Optimization (SRPO), a unified on-policy framework that routes correct samples to GRPO's reward-aligned reinforcement and failed samples to SDPO's targeted logit-level correction. SRPO further incorporates an entropy-aware dynamic weighting mechanism to suppress high-entropy, unreliable distillation targets while emphasizing confident ones. Evaluated across five benchmarks and two model scales, SRPO achieves both the rapid early improvement of SDPO and the long-horizon stability of GRPO. It consistently surpasses the peak performance of both baselines, raising the five-benchmark average on Qwen3-8B by 3.4% over GRPO and 6.3% over SDPO, while simultaneously yielding moderate response lengths and lowering per-step compute cost by up to 17.2%.

📄 PDF Abstract BibTeX arXiv:2604.02288

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement Learning

Similar Papers 제목 키워드 기반

I-SDPO: Instance-Level Adaptive Self-Distillation Policy Optimization

2026-08-13 · Yubo Zhang, Xinhong Ma, Zezhong Tan, Ziqiang Dong arxiv

Group Relative Policy Optimization (GRPO) learns from reward differences within a rollout group, but receives no useful relative signal when every sampled response is incorrect. Privileged self-distillation can fill this…

Self-Distilled Policy Gradient

2026-06-02 · Yifeng Liu, Shiyuan Zhang, Yifan Zhang, Quanquan Gu arxiv

On-policy self-distillation, where a language model conditions on privileged context to supervise its own generations, is a promising source of dense supervision for sparse-reward reinforcement learning. Actually, it can…

Reinforcement Learning

When Models Judge Themselves: Unsupervised Self-Evolution for Multimodal Reasoning

2026-03-22 · Zhengxian Wu, Kai Shi, Chuanrui Zhang, Zirui Liao 외 arxiv

Recent progress in multimodal large language models has led to strong performance on reasoning tasks, but these improvements largely rely on high-quality annotated data or teacher-model distillation, both of which are co…

Mathematical ReasoningMultimodal Reasoning

FIRE-VLA: Failure-Informed Self-Evolution for Vision-Language-Action Models in Autonomous Driving

2026-08-13 · Hao Dou arxiv

Reinforcement learning improves autonomous-driving vision-language-action (VLA) models by evaluating trajectories sampled from the current policy. Group relative policy optimization (GRPO) learns from reward differences …

Reinforcement LearningAutonomous Driving

CAST: Non-Privileged Clipped Asymmetric Self-Teaching with Advantage Flipping for GRPO

2026-05-29 · Yang Li, Gongle Xue, Yijia Guo, Yuheng Yuan 외 arxiv

Reinforcement learning with verifiable rewards (RLVR), especially Group Relative Policy Optimization (GRPO), has been widely used to improve reasoning in large language models. However, outcome-level rewards provide only…

Reinforcement LearningMathematical Reasoning