Diagnosing On-Policy Self-Distillation for Reasoning Language Models

arXiv:2609.39118 · cs.CL · Submitted 2026-09-30 · Read on arXiv

cs.CL

Submitted: 2026-09-30

Updated: 2026-09-30

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• GRPO — 60× in this paper · explained in 6 episodes
• Benchmark — 28× in this paper · explained in 2 episodes
• Distillation — 28× in this paper · explained in 2 episodes
• Accuracy — 22× in this paper · explained in 1 episode
• Self-distillation — 14× in this paper · explained in 6 episodes
• Trace — 12× in this paper · explained in 2 episodes
• Ablation — 11× in this paper · explained in 1 episode
• Actor — 11× in this paper · explained in 3 episodes
• On-Policy Distillation — 11× in this paper · explained in 3 episodes
• ON-POLICY SELF-DISTILLATION — 7× in this paper
• Robustness — 6× in this paper · explained in 5 episodes
• Reinforcement Learning — 5× in this paper · explained in 4 episodes
• Credit assignment — 4× in this paper · explained in 3 episodes
• Decode — 3× in this paper · explained in 2 episodes
• Transformer — 3× in this paper · explained in 2 episodes
• Verification — 3× in this paper · explained in 2 episodes
• Alignment — 2× in this paper · explained in 1 episode
• Core — 2× in this paper · explained in 2 episodes
• Generalization — 2× in this paper · explained in 13 episodes
• On-policy Self-Distillation (OPSD) — 2× in this paper · explained in 2 episodes
• Adaptation — 1× in this paper · explained in 1 episode
• Hallucination — 1× in this paper · explained in 3 episodes
• Large Language Models (LLMs) — 1× in this paper · explained in 6 episodes
• Regularization — 1× in this paper · explained in 3 episodes
• Test-time Scaling — 1× in this paper · explained in 2 episodes
• Tokenization — 1× in this paper · explained in 2 episodes
• Verifiable Rewards — 1× in this paper · explained in 2 episodes

Sources

• On-Policy Distillation of Language Models: Learning from Self-Generated Mistakes
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
• Falcon-H1R: Pushing the Reasoning Frontiers with a Hybrid Model for Efficient Test-Time Scaling
• LoRA: Low-Rank Adaptation of Large Language Models
• Rethinking On-Policy Self-Distillation for Thinking Models
• Why Does Self-Distillation (Sometimes) Degrade the Reasoning Capability of LLMs?
• Kimi K3: Open Frontier Intelligence
• Efficient Memory Management for Large Language Model Serving with PagedAttention
• Unifying Group-Relative and Self-Distillation Policy Optimization via Sample Routing
• Localizing Credit at the Divergence: Path-Conditioned Self-Distillation for LLM Reasoning
• When Are Teacher Tokens Reliable? Position-Weighted On-Policy Self-Distillation for Reasoning
• Understanding R1-Zero-Like Training: A Critical Perspective
• MiMo: Unlocking the Reasoning Potential of Language Model -- From Pretraining to Posttraining
• Decoupled Weight Decay Regularization
• Demystifying OPD: Length Inflation and Stabilization Strategies for Large Language Models
• AIMO-2 Winning Solution: Building State-of-the-Art Mathematical Reasoning Models with OpenMathReasoning dataset
• Olmo 3
• RLCSD: Reinforcement Learning with Contrastive On-Policy Self-Distillation
• Qwen3 Technical Report
• ZeRO: Memory Optimizations Toward Training Trillion Parameter Models

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering