---
id: 20261018-T0-05
title: "技能与策略共同进化会“一起进化、各自失败”：自改进Agent的技能-策略协同陷阱"
title_en: "Evolving Together, Failing Apart: The Skill-Policy Co-evolution Trap in Self-Improving Agents"
url: https://ai.daily.yangsir.net/daily/20261018-T0-05
issue_date: 2026-10-18
publish_date: 2026-10-17T04:00:00.000Z
category: research
source_name: "arXiv cs.AI"
source_url: https://arxiv.org/abs/2610.20845
---

# 技能与策略共同进化会“一起进化、各自失败”：自改进Agent的技能-策略协同陷阱

论文研究自改进代理中的技能-策略协同进化：训练过程中策略模型与技能库相互提升。作者指出当前评估方式存在盲区——只在共同进化结束时的最终技能库上评测策略，因此看不到协同过程本身带来的问题，并将其称为技能-策略协同进化陷阱。这对自改进代理的评测方法提出质疑：如果评估只用最终状态，训练过程中的退化或不匹配可能被掩盖，实际部署时的鲁棒性会被高估。

## English Version

**Evolving Together, Failing Apart: The Skill-Policy Co-evolution Trap in Self-Improving Agents**

The paper examines skill-policy co-evolution in self-improving agents, where an agent's policy model and its skill bank improve each other during training. The authors argue current evaluations have a blind spot: they score the co-evolved policy only on the final skill bank it trained with, so problems introduced by the co-evolution process itself go unseen. They call this the skill-policy co-evolution trap. The implication is a challenge to evaluation practice: scoring only the end state can mask degradation or mismatch during training, leading to overestimated robustness in deployment.

---

**来源**：[arXiv cs.AI](https://arxiv.org/abs/2610.20845)

**详情页**：https://ai.daily.yangsir.net/daily/20261018-T0-05

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*