---
id: 20261027-T0-10
title: "代理犯错后总忘？自我定向修复让教训跨会话留存"
title_en: "Self-Directed Repair Turns Agent Mistakes Into Lessons That Persist Across Sessions"
url: https://ai.daily.yangsir.net/daily/20261027-T0-10
issue_date: 2026-10-27
publish_date: 2026-10-26T04:00:00.000Z
category: research
source_name: "arXiv cs.AI"
source_url: https://arxiv.org/abs/2610.31846
---

# 代理犯错后总忘？自我定向修复让教训跨会话留存

当前智能代理从错误中恢复只是暂时的：上下文内的反思在跨会话时容易被遗忘，还要和有限的上下文窗口争空间。这项研究提出自我定向修复方法，把代理犯过的错误转化为可长期保留的经验，让智能体在后续任务中不再重蹈覆辙。对需要长期运行的代理系统来说，这解决的是"同一个坑反复踩"的问题。论文为arXiv预印本。

## English Version

**Self-Directed Repair Turns Agent Mistakes Into Lessons That Persist Across Sessions**

Today's agents recover from mistakes only temporarily: in-context reflections are easily forgotten across sessions and compete for limited context space. This work proposes self-directed repair, which converts agent mistakes into experience that persists, so agents avoid repeating the same errors in later tasks. For long-running agent systems, it addresses the problem of falling into the same trap repeatedly. The paper is an arXiv preprint.

---

**来源**：[arXiv cs.AI](https://arxiv.org/abs/2610.31846)

**详情页**：https://ai.daily.yangsir.net/daily/20261027-T0-10

---

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