Unbiased Top- k Estimation for On-Policy Distillation

arXiv:2609.34447 · cs.CL, cs.LG, stat.ML · Submitted 2026-09-28 · Read on arXiv

cs.CL, cs.LG, stat.ML

Submitted: 2026-09-28

Updated: 2026-10-01

Terminology

• Distillation — 41× in this paper · explained in 2 episodes
• Accuracy — 35× in this paper · explained in 1 episode
• On-Policy Distillation — 20× in this paper · explained in 3 episodes
• GRPO — 12× in this paper · explained in 6 episodes
• Robustness — 9× in this paper · explained in 5 episodes
• Benchmark — 5× in this paper · explained in 2 episodes
• Exposure Bias — 5× in this paper · explained in 2 episodes
• Ablation — 4× in this paper · explained in 1 episode
• Knowledge Distillation — 2× in this paper · explained in 11 episodes
• Reinforcement Learning — 2× in this paper · explained in 4 episodes
• Self-Distillation — 2× in this paper · explained in 7 episodes
• Tail-Corrected Top-k On-Policy Distillation — 2× in this paper
• Adaptive Weighting — 1× in this paper · explained in 1 episode
• Aggregation — 1× in this paper · explained in 1 episode
• Alignment — 1× in this paper · explained in 1 episode
• Consistency — 1× in this paper · explained in 3 episodes
• Continual Learning — 1× in this paper · explained in 2 episodes
• Core — 1× in this paper · explained in 2 episodes
• Failure Modes — 1× in this paper · explained in 3 episodes
• Large Language Model (LLM) — 1× in this paper · explained in 3 episodes
• Multimodal Reasoning — 1× in this paper · explained in 5 episodes
• Pruning — 1× in this paper · explained in 4 episodes
• Teacher-Student Gap — 1× in this paper · explained in 2 episodes
• Thinking mode — 1× in this paper · explained in 2 episodes
• Verification — 1× in this paper · explained in 2 episodes

Sources

• Process Reinforcement through Implicit Rewards
• Revisiting On-Policy Distillation: Empirical Failure Modes and Simple Fixes
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
• Skywork Open Reasoner 1 Technical Report
• Distilling the Knowledge in a Neural Network
• Uni-OPD: Unifying On-Policy Distillation with a Dual-Perspective Recipe
• Reinforcement Learning via Self-Distillation
• DistiLLM-2: A Contrastive Approach Boosts the Distillation of LLMs
• Rethinking On-Policy Distillation of Large Language Models: Phenomenology, Mechanism, and Recipe
• KL for a KL: On-Policy Distillation with Control Variate Baseline
• Self-Distillation Enables Continual Learning
• HybridFlow: A Flexible and Efficient RLHF Framework
• A Survey of On-Policy Distillation for Large Language Models
• Not All Disagreement Is Learnable: Token Teachability in On-Policy Distillation
• Lightning OPD: Efficient Post-Training for Large Reasoning Models with Offline On-Policy Distillation
• MiMo-V2-Flash Technical Report
• On the Position Bias of On-Policy Distillation
• Escaping the KL Agreement Trap in On-Policy Distillation
• DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence
• A Survey on Knowledge Distillation of Large Language 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