RL Starts before RL: On Policy Distillation for Better Reinforcement Learning

arXiv:2609.28145 · cs.LG · Submitted 2026-09-23 · Read on arXiv

cs.LG

Submitted: 2026-09-23

Updated: 2026-09-23

Terminology

• Distillation — 47× in this paper · explained in 2 episodes
• Benchmark — 42× in this paper · explained in 2 episodes
• Alignment — 29× in this paper · explained in 1 episode
• Accuracy — 18× in this paper · explained in 1 episode
• Reinforcement Learning — 10× in this paper · explained in 4 episodes
• On-Policy Distillation — 8× in this paper · explained in 3 episodes
• Aggregation — 5× in this paper · explained in 1 episode
• Fine-Tuning — 5× in this paper · explained in 5 episodes
• Verifier — 5× in this paper · explained in 2 episodes
• Lexical Diversity — 4× in this paper · explained in 2 episodes
• GRPO — 3× in this paper · explained in 6 episodes
• Actor — 2× in this paper · explained in 3 episodes
• Knowledge Distillation — 2× in this paper · explained in 11 episodes
• Reinforcement Learning (RL) — 2× in this paper · explained in 25 episodes
• Ablation — 1× in this paper · explained in 1 episode
• Failure Modes — 1× in this paper · explained in 3 episodes
• Multimodal Reasoning — 1× in this paper · explained in 5 episodes
• Objective Function — 1× in this paper · explained in 2 episodes
• Supervised Fine-Tuning (SFT) — 1× in this paper · explained in 8 episodes
• Verifiable Rewards — 1× in this paper · explained in 2 episodes

Sources

• Qwen3-VL Technical Report
• Behavior Injection: Preparing Language Models for Reinforcement Learning
• Evaluating Large Language Models Trained on Code
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
• Revisiting On-Policy Distillation: Empirical Failure Modes and Simple Fixes
• OlympiadBench: A Challenging Benchmark for Promoting AGI with Olympiad-Level Bilingual Multimodal Scientific Problems
• DeepMath-103K: A Large-Scale, Challenging, Decontaminated, and Verifiable Mathematical Dataset for Advancing Reasoning
• Distilling the Knowledge in a Neural Network
• Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes
• Vision-R1: Incentivizing Reasoning Capability in Multimodal Large Language Models
• Entropy-Aware On-Policy Distillation of Language Models
• Quagmires in SFT-RL Post-Training: When High SFT Scores Mislead and What to Use Instead
• Solving Quantitative Reasoning Problems with Language Models
• Rethinking On-Policy Distillation of Large Language Models: Phenomenology, Mechanism, and Recipe
• ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models
• TailSFT: Filtered Fine-Tuning Improves Post-Training Performance
• Offline Exploration-Aware Fine-Tuning for Long-Chain Mathematical Reasoning
• DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
• OpenMathInstruct-2: Accelerating AI for Math with Massive Open-Source Instruction Data
• Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning

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: Local-Consistency Optimisation for Bayesian Neural Networks
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks