OPSD Compresses What RLVR Teaches: A Post-RL Compaction Stage for Reasoning Models

• URL: http://arxiv.org/abs/2605.06188v1
• Date: Thu, 07 May 2026 13:04:34 GMT
• Title: OPSD Compresses What RLVR Teaches: A Post-RL Compaction Stage for Reasoning Models
• Abstract summary: On-Policy Self-Distillation (OPSD) has recently emerged as an alternative to Reinforcement Learning with Verifiable Rewards (RLVR)<n>We show that in thinking-enabled mathematical reasoning, OPSD behaves most reliably as a compression mechanism rather than a correction mechanism.
• Score: 12.17078443900398
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: On-Policy Self-Distillation (OPSD) has recently emerged as an alternative to Reinforcement Learning with Verifiable Rewards (RLVR), promising higher accuracy and shorter responses through token-level credit assignment from a self-teacher conditioned on privileged context. However, this promise does not carry over to thinking-enabled mathematical reasoning, where reported accuracy gains shrink and sometimes turn negative. We hypothesize that hindsight supervision can specify better token-level alternatives in short thinking-disabled outputs, but in long thinking-enabled traces it more readily identifies redundancy than supplies better replacements. To test this, we applied OPSD separately to correct and incorrect rollout groups, so that compression and correction can be observed in isolation. Our results show that in thinking-enabled mathematical reasoning, OPSD behaves most reliably as a compression mechanism rather than a correction mechanism: training only on correct rollouts preserves accuracy while substantially shortening responses, whereas training only on incorrect rollouts damages accuracy. In light of these findings, we propose a revised post-training pipeline for thinking-enabled mathematical reasoning: SFT then RLVR then OPSD.

Related papers

• SVR: Self-Verifying Refinement via Joint Verdict-Confidence Reinforcement Learning for Adaptive Test-Time Compute [62.3458279176813]
  Self-Verifying Refinement is an oracle-free multi-turn reinforcement learning framework.<n>It learns to use self-verification as a compute-control policy.<n>It achieves a macro-average accuracy of 0.563 with only 2.99 inference turns on average.
  arXiv  Detail & Related papers  (2026-07-30T16:20:58Z)
• H$^2$SD: Hybrid Hindsight Self-Distillation [61.71131241473028]
  Reinforcement learning with verifiable rewards (RLVR) provides reliable outcome supervision for language model reasoning.<n>Existing self-distillation methods add a privileged teacher but typically assign it a fixed role.<n>We introduce Hybrid Hindsight Self-Distillation ($mathrmH2mathrmSD$), which jointly adapts teacher context and update strategy to trajectory correctness.
  arXiv  Detail & Related papers  (2026-07-21T10:47:27Z)
• Rubric-Guided Self-Distillation: Post-Training Without Rubric Verifiers [25.040339775169883]
  We propose a verifier-free training method in which the base policy, conditioned on the rubric, serves as the teacher for the unconditioned student.<n>Across Qwen-2.5 (3B, 7B) and Qwen3-Thinking (4B, 8B) models on medical and science domains, RGSD achieves satisfaction comparable to judge-based rubricO.
  arXiv  Detail & Related papers  (2026-06-10T17:53:19Z)
• CAST: Non-Privileged Clipped Asymmetric Self-Teaching with Advantage Flipping for GRPO [9.443660785229719]
  Reinforcement learning with verifiable rewards (RLVR) has been widely used to improve reasoning in large language models.<n>This work proposes CAST, an answer-free self-distillation method for GRPO-style RLVR.
  arXiv  Detail & Related papers  (2026-05-29T13:21:30Z)
• Adaptive Negative Reinforcement for LLM Reasoning:Dynamically Balancing Correction and Diversity in RLVR [6.291433683451589]
  Reinforcement learning with verifiable rewards (RLVR) has become a highly effective method for improving the reasoning abilities of Large Language Models (LLMs)<n>Recent research shows that Negative Sample Reinforcement (NSR) can match or even exceed the performance of more complex frameworks like PPO and GRPO.<n>We propose two extensions to the NSR framework: Adaptive Negative Sample Reinforcement and Confidence-Weighted Negative Reinforcement.
  arXiv  Detail & Related papers  (2026-05-08T02:13:33Z)
• Correct Answers from Sound Reasoning: Verifiable Process Supervision for Language Models [94.68358825189738]
  We propose a post-training framework for verifiable domains that jointly optimize prediction accuracy and reasoning quality.<n>We evaluate VPS on chess, a controlled testbed where reasoning steps can be deterministically verified against engine signals.<n>VPS preserves accuracy while significantly improving reasoning quality, reducing win-rate error by up to 30% and restoring consistency to near saturation.
  arXiv  Detail & Related papers  (2026-04-03T15:19:46Z)
• What If Consensus Lies? Selective-Complementary Reinforcement Learning at Test Time [57.533031432715084]
  Test-Time Reinforcement Learning (TTRL) enables Large Language Models (LLMs) to enhance reasoning capabilities on unlabeled test streams.<n>Existing TTRL methods rely exclusively on positive pseudo-labeling strategies.<n>We propose SCRL, a robust test-time reinforcement learning framework that effectively mitigates label noise amplification.
  arXiv  Detail & Related papers  (2026-03-20T11:47:12Z)
• Overconfident Errors Need Stronger Correction: Asymmetric Confidence Penalties for Reinforcement Learning [17.384089089363382]
  We identify a root cause that existing methods overlook: the uniform penalization of errors.<n>Current approaches treat all incorrect rollouts within a group identically.<n>We propose the Asymmetric Confidence-aware Error Penalty (ACE)
  arXiv  Detail & Related papers  (2026-02-24T22:46:43Z)
• Constraint-Rectified Training for Efficient Chain-of-Thought [60.52883907721588]
  Chain-of-Thought (CoT) has significantly enhanced the reasoning capabilities of Large Language Models (LLMs)<n>While longer reasoning traces can improve answer quality and unlock abilities such as self-correction, they also incur high inference costs and often introduce redundant steps, known as overthinking.<n>Recent research seeks to develop efficient reasoning strategies that balance reasoning length and accuracy.
  arXiv  Detail & Related papers  (2026-02-13T02:13:45Z)
• Beyond Correctness: Learning Robust Reasoning via Transfer [51.403609251508904]
  We adopt a simple philosophical view, robust reasoning should remain useful beyond the mind that produced it.<n>We introduce Reinforcement Learning with Transferable Reward, which operationalizes robustness via transfer reward.<n>Our approach improves sampling consistency while improving final answer accuracy, and it reaches comparable performance in substantially fewer training steps.
  arXiv  Detail & Related papers  (2026-02-09T10:41:44Z)
• Step Potential Advantage Estimation: Harnessing Intermediate Confidence and Correctness for Efficient Mathematical Reasoning [25.562101968892833]
  Reinforcement Learning with Verifiable Rewards (RLVR) elicits long chain-of-thought reasoning in large language models (LLMs)<n>Existing approaches improve RLVR via token-level entropy or sequence-level length control, but lack a semantically grounded, step-level measure of reasoning progress.<n>We propose Step Potential Advantage Estimation (SPAE), a fine-grained credit assignment method that amplifies potential gains, penalizes potential drops, and applies penalty after potential saturates to encourage timely termination.
  arXiv  Detail & Related papers  (2026-01-07T11:36:01Z)
• Sandwich Reasoning: An Answer-Reasoning-Answer Approach for Low-Latency Query Correction [14.620900910188382]
  Chain-of-Thought (CoT) reasoning improves accuracy but incurs prohibitive latency for real-time query correction.<n>We propose Sandwich Reasoning (SandwichR), a novel approach that explicitly aligns a fast initial answer with post-hoc reasoning.<n>We show that SandwichR achieves SOTA accuracy comparable to standard CoT while delivering a 40-70% latency reduction.
  arXiv  Detail & Related papers  (2026-01-07T07:52:30Z)
• Beyond Binary Rewards: Training LMs to Reason About Their Uncertainty [59.97939500426759]
  This paper describes RLCR, an approach to training reasoning models that jointly improves accuracy and confidence estimation.<n>We show that across diverse datasets, RLCR substantially improves calibration with no loss in accuracy.<n>We also demonstrate that verbalized confidence can be leveraged at test time to improve accuracy and calibration.
  arXiv  Detail & Related papers  (2025-07-22T17:56:01Z)
• The Surprising Effectiveness of Negative Reinforcement in LLM Reasoning [37.13807960501503]
  Reinforcement learning with verifiable rewards (RLVR) is a promising approach for training language models (LMs)<n>We decompose the learning signal into reinforcing correct responses and penalizing incorrect ones, referred to as Positive and Negative Sample Reinforcement (PSR and NSR)<n>We show that NSR works by suppressing incorrect generations and redistributing probability mass toward other plausible candidates, guided by the model's prior beliefs.
  arXiv  Detail & Related papers  (2025-06-02T06:10:54Z)
• SEAL: Steerable Reasoning Calibration of Large Language Models for Free [58.931194824519935]
  Large Language Models (LLMs) have demonstrated compelling capabilities for complex reasoning tasks via the extended chain-of-thought (CoT) reasoning mechanism.<n>Recent studies reveal substantial redundancy in the CoT reasoning traces, which negatively impacts model performance.<n>We introduce SEAL, a training-free approach that seamlessly calibrates the CoT process, improving accuracy while demonstrating significant efficiency gains.
  arXiv  Detail & Related papers  (2025-04-07T02:42:07Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.