paper-with-me

홈 › Papers

OPID: On-Policy Skill Distillation for Agentic Reinforcement Learning

2026-06-25 · Shuo Yang, Jinyang Wu, Zhengxi Lu, Yuhao Shen, Fan Zhang, Lang Feng, Shuai Zhang, Haoran Luo, Zheng Lian, Zhengqi Wen, Jianhua Tao arxiv

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 introducing dense, distribution-matched hindsight supervision. Experiments on ALFWorld, WebShop and Search-based QA demonstrate that OPID generally improves agent performance, sample efficiency, and robustness over outcome-only RL and existing skill-distillation baselines. Our code is available at https://github.com/jinyangwu/OPID/tree/main.

📄 PDF Abstract BibTeX arXiv:2606.26790

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement Learning

Similar Papers 제목 키워드 기반

UCOB: Learning to Utilize and Evolve Agentic Skills via Credit-Aware On-Policy Bidirectional Self-Distillation

2026-06-28 · Songjun Tu, Chengdong Xu, Qichao Zhang, Yiwen Ma 외 arxiv

Skill memories can improve agentic reinforcement learning by reusing past experience as textual guidance, but retrieved skills are not oracular: they may help in one state while misleading the same policy in another. Thi…

Reinforcement Learning

RetireOPD: Self-Retiring On-Policy Distillation for Agentic Reinforcement Learning

2026-09-17 · Yan Yu, Zhengxi Lu, Yizhou Liu, Yichen Pan 외 hf

Multi-turn agents trained with reinforcement learning (RL) receive a single scalar reward per trajectory, which motivates self on-policy distillation (OPD) to supply dense token-level supervision from a self-teacher with…

Reinforcement Learning

SEED: Self-Evolving On-Policy Distillation for Agentic Reinforcement Learning

2026-07-16 · Jinyang Wu, Shuo Yang, Zhengxi Lu, Fan Zhang 외 arxiv

Large language models are increasingly trained as interactive agents for long-horizon tasks involving multi-turn interaction, tool use, and environment feedback. Outcome-based reinforcement learning (RL) provides a pract…

Reinforcement LearningDecision Making

Group-Reflective Self-Distillation for Agentic Reinforcement Learning

2026-07-30 · Binbin Zheng, Zijun Xie, Guanqun Zhao, Enlei Gong 외 arxiv

Reinforcement learning with verifiable rewards (RLVR) is effective for training large language model agents. However, terminal rewards provide only coarse trajectory-level supervision, leaving successful behaviors, recur…

Reinforcement Learning

Self-Distilled Agentic Reinforcement Learning

2026-05-14 · Zhengxi Lu, Zhiyuan Yao, Zhuowen Han, Zi-Han Wang 외 arxiv

Reinforcement learning (RL) has emerged as a central paradigm for post-training LLM agents, yet its trajectory-level reward signal provides only coarse supervision for long-horizon interaction. On-Policy Self-Distillatio…

Reinforcement Learning