Episode
OPID: On-Policy Skill Distillation for Agentic Reinforcement Learning
- Podcast
- Daily Paper Cast
- Published
- Jun 28, 2026
- Duration seconds
- 1432
- Processing state
not_requested- Canonical source
- https://share.transistor.fm/s/a6ccf440
Actions
POST https://stenobird.com/v1/public/podcasts/daily-paper-cast-7079649/episodes/opid-on-policy-skill-distillation-for-agentic-reinforcement-learning/transcription-requests
Idempotently request low-priority transcript generation for this episode.GET https://stenobird.com/podcast/daily-paper-cast-7079649/opid-on-policy-skill-distillation-for-agentic-reinforcement-learning.md
Read the agent-friendly Markdown representation of this episode resource.
Summary
🤗 Upvotes: 40 | cs.CL Authors: Shuo Yang, Jinyang Wu, Zhengxi Lu, Yuhao Shen, Fan Zhang, Lang Feng, Shuai Zhang, Haoran Luo, Zheng Lian, Zhengqi Wen, Jianhua Tao Title: OPID: On-Policy Skill Distillation for Agentic Reinforcement Learning Arxiv: http://arxiv.org/abs/2606.26790v1 Abstract: Outcome-based reinforcement learning provides a stable optimization backbone for language agents, but its sparse trajectory-level rewards provide little guidance on which intermediate decisions should be reinforced or suppressed. On-policy self-distillation offers dense token-level supervision, yet existing skill-conditioned variants often rely on external skill memories or retrieved privileged context, which are costly to maintain and can be mismatched with the state distribution induced by the current policy in multi-turn interaction. We propose \textbf{OPID} (\textbf{O}n-\textbf{P}olicy Sk\textbf{i}ll \textbf{D}istillation), a framework that extracts skill supervision directly from completed on-policy trajectories. OPID represents trajectory hindsight as hierarchical skills: episode-level skills capture global workflows or failure-avoidance rules, while step-level skills capture local decision knowledge at critical timesteps. A critical-first routing mechanism uses step-level skills when critical decisions are identified and falls back to episode-level skills as default guidance otherwise. The selected skill is injected into the interaction history, allowing the old policy to re-score the same sampled response under both original and skill-augmented contexts. The resulting log-probability shift yields a token-level self-distillation advantage, which is combined with the outcome advantage for policy optimization. OPID thus preserves RL as the primary training objective while introdu…