Paper Detail
Zhilin Wang, Han Song, Runzhe Zhan, Jusen Du, Jiacheng Chen, Tianle Li, Qingyu Yin, Yulun Wu, Zhennan Shen, Tong Zhu, Yanshu Li, Guanjie Chen, Derek F. Wong, Yafu Li, Yu Cheng, Yang Yang
Autonomous agents are increasingly expected to improve executable policies through feedback, yet existing evaluations often collapse this process into a final score or confound it with open-ended software-engineering progress. We introduce Autonomous Policy Evolution, a controlled evaluation setting in which a harness-model agent repeatedly edits an executable policy system under a fixed interaction budget. We instantiate this setting in EvoPolicyGym, a benchmark built from compact interactive RL environments that evaluates how agents iteratively improve explored policies. On the EvoPolicyGym suite, GPT-5.5 achieves the strongest aggregate rank score and top-two performance on all 16 environments. Beyond leaderboard results, EvoPolicyGym also provides trajectory-level diagnostics that distinguish how agents allocate budget, convert feedback into parametric tuning. These analyses show that strong autonomous policy evolution depends not only on isolated task wins, but on discovering task-appropriate mechanisms and refining policies under bounded feedback.
No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.
No ranking explanation is available yet.
No tags.
@misc{wang2026evopolicygym,
title = {EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments},
author = {Zhilin Wang and Han Song and Runzhe Zhan and Jusen Du and Jiacheng Chen and Tianle Li and Qingyu Yin and Yulun Wu and Zhennan Shen and Tong Zhu and Yanshu Li and Guanjie Chen and Derek F. Wong and Yafu Li and Yu Cheng and Yang Yang},
year = {2026},
abstract = {Autonomous agents are increasingly expected to improve executable policies through feedback, yet existing evaluations often collapse this process into a final score or confound it with open-ended software-engineering progress. We introduce Autonomous Policy Evolution, a controlled evaluation setting in which a harness-model agent repeatedly edits an executable policy system under a fixed interaction budget. We instantiate this setting in EvoPolicyGym, a benchmark built from compact interactive R},
url = {https://huggingface.co/papers/2607.02440},
keywords = {autonomous agents, executable policies, policy evolution, reinforcement learning, trajectory-level diagnostics, parametric tuning, task-appropriate mechanisms, huggingface daily},
eprint = {2607.02440},
archiveprefix = {arXiv},
}
{}