Episode
SEED: Self-Evolving On-Policy Distillation for Agentic Reinforcement Learning
- Podcast
- Daily Paper Cast
- Published
- Jul 18, 2026
- Duration seconds
- 1156
- Processing state
not_requested- Canonical source
- https://share.transistor.fm/s/170dbcfd
Actions
POST https://stenobird.com/v1/public/podcasts/daily-paper-cast-7079649/episodes/seed-self-evolving-on-policy-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/seed-self-evolving-on-policy-distillation-for-agentic-reinforcement-learning.md
Read the agent-friendly Markdown representation of this episode resource.
Summary
🤗 Upvotes: 71 | cs.CL Authors: Jinyang Wu, Shuo Yang, Zhengxi Lu, Fan Zhang, Yuhao Shen, Lang Feng, Haoran Luo, Zheng Lian, Shuai Zhang, Zhengqi Wen, Jianhua Tao Title: SEED: Self-Evolving On-Policy Distillation for Agentic Reinforcement Learning Arxiv: http://arxiv.org/abs/2607.14777v1 Abstract: 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 practical optimization paradigm, but its sparse trajectory-level rewards offer limited guidance on intermediate decisions, leaving a supervision gap between episode-level outcomes and token-level policy learning. We propose SEED (SElf-Evolving On-Policy Distillation), a self-evolving framework that converts completed on-policy trajectories into training-time hindsight skills and distills their behavioral effect back into the policy model. SEED first fine-tunes the policy to analyze completed trajectories and generate natural-language skills that capture reusable workflows, decisive observations, or failure-avoidance rules. During RL, the current policy both collects trajectories and serves as the analyzer that extracts hindsight skills from them. Policy updates therefore improve subsequent decision making and skill analysis together, allowing hindsight supervision to evolve with the policy. SEED then re-scores the sampled actions under ordinary and skill-augmented contexts, converting the skill-induced probability shift into a dense token-level on-policy distillation signal. This signal is jointly optimized with outcome-based RL, keeping the auxiliary supervision aligned with the current trajectory distribution. Extensive experiments on text-based and vision-based a…