When Sparse Reward Meets Dense Distillation: Training Dynamics of On-Policy Distillation

arXiv:2609.34849 · cs.LG · Submitted 2026-09-28 · Read on arXiv

cs.LG

Submitted: 2026-09-28

Updated: 2026-09-28

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

Terminology

• Distillation — 2× in this paper · explained in 2 episodes
• On-Policy Distillation — 1× in this paper · explained in 3 episodes

Sources

• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
• HDPO: Hybrid Distillation Policy Optimization via Privileged Self-Distillation
• Reinforcement Learning via Self-Distillation
• Direct Preference Knowledge Distillation for Large Language Models
• Understanding R1-Zero-Like Training: A Critical Perspective
• NTKMTL: Mitigating Task Imbalance in Multi-Task Learning from Neural Tangent Kernel Perspective
• Proximal Policy Optimization Algorithms
• DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
• Self-Distilled RLVR
• DAPO: An Open-Source LLM Reinforcement Learning System at Scale
• Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Models

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