E squared-OPSD: Taming Entropy Overshoot in On-Policy Self-Distillation

arXiv:2610.05048 · cs.LG, cs.AI · Submitted 2026-10-04 · Read on arXiv

cs.LG, cs.AI

Submitted: 2026-10-04

Updated: 2026-10-15

Terminology

• Distillation — 1× in this paper · explained in 2 episodes
• Self-Distillation — 1× in this paper · explained in 7 episodes

Sources

• A General Language Assistant as a Laboratory for Alignment
• XRPO: Pushing the limits of GRPO with Targeted Exploration and Exploitation
• Evaluating Large Language Models Trained on Code
• The Entropy Mechanism of Reinforcement Learning for Reasoning Language Models
• DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence
• GLM-5: from Vibe Coding to Agentic Engineering
• MiniLLM: On-Policy Distillation of Large Language Models
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
• Adaptive Teacher Exposure for Self-Distillation in LLM Reasoning
• Rethinking Entropy Interventions in RLVR: An Entropy Change Perspective
• Privileged, but Biased: How PI-Conditioned Teachers Break Self-Distillation
• Self-Distillation Zero: Self-Revision Turns Binary Rewards into Dense Supervision
• DeepMath-103K: A Large-Scale, Challenging, Decontaminated, and Verifiable Mathematical Dataset for Advancing Reasoning
• Reinforcement Learning via Self-Distillation
• Distill Skills into Weights, Not Prompts: Abstract Skills as Privileged Signals for On-Policy Self-Distillation
• Entropy-Aware On-Policy Distillation of Language Models
• Masked Self-Distillation: Internalizing the Chain-of-Thought in Language Models
• Rethinking On-Policy Self-Distillation for Thinking Models
• Respecting Self-Uncertainty in On-Policy Self-Distillation for Efficient LLM Reasoning
• Why Does Self-Distillation (Sometimes) Degrade the Reasoning Capability of LLMs?

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks