On the Off-Policy Teacher in On-Policy Distillation
cs.LG, cs.AI, cs.CL
Submitted: 2026-09-29
Updated: 2026-09-29
Code: https://github.com/verl-project/verl
Terminology
Sources
- Your Teacher Can't Help You Here: Combating Supervision Fidelity Decay in On-Policy Distillation
- Rethinking On-Policy Distillation of Large Language Models: Phenomenology, Mechanism, and Recipe
- Trust Region On-Policy Distillation
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- Qwen3 Technical Report
- Skywork Open Reasoner 1 Technical Report
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- Prune-OPD: Efficient and Reliable On-Policy Distillation for Long-Horizon Reasoning
- Pass the Baton: Trajectory-Relayed On-Policy Distillation
- Evaluating Large Language Models Trained on Code
- Program Synthesis with Large Language Models
- Are Full Rollouts Necessary for On-Policy Distillation?
- On the Position Bias of On-Policy Distillation
- SOD: Step-wise On-policy Distillation for Small Language Model Agents
- Reward-Gated On-Policy Distillation
- SPOT: Sparse Probing and Outcome Calibration for On-Policy Distillation
- Backtracking When It Strays: Mitigating Dual Exposure Biases in LLM Reasoning Distillation
- Trajectory-Refined Distillation
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