原帖引用 · AI 翻译
标题：“别忘了刚学到的东西：自主修复将智能体的错误转化为持久的经验教训”
普林斯顿大学和 Meta 的一篇关于 AI 智能体持续改进的不错的新论文
研究发现，如果将每次修复保存为 LoRA 适配器，并在类似任务中重复使用，智能体就会更少重复犯错。
智能体可以在一次会话中纠正错误，却常常在下一次会话中重犯。上下文中的笔记会被遗忘，而对整个模型进行微调成本高昂，还可能影响其他技能。
借助 Self-Direction，任务失败的智能体会反思自己的运行过程，并将学到的经验训练成一个大小约为模型 0.02% 的 LoRA 适配器。之后，当智能体在类似任务中失败时，路由器会加载最匹配的适配器。
使用 Qwen3.6-27B 时，GAIA 上的任务成功率从 38.8% 升至 55.8%，优于上下文内反思和 Skills.md 文件。在先前的 185 次错误中，到第 3 次遇到相同情况时，67.3% 的修复仍然有效，而直接重试的这一比例为 42.9%。
如果你的智能体不断重犯同样的错误，就把每次修复保存为它们可以重新加载的小型适配器，而不只是存为提示词笔记。
原文
Nice new paper on continuous improvement of AI agents from Princeton and Meta Finds that agents repeat fewer mistakes when each fix is saved as a LoRA adapter and reused on similar tasks. Agents can fix a mistake within a session, then often repeat it in the next one. Notes in context get forgotten, and full fine-tuning is costly and can disrupt other skills. With Self-Direction, a failing agent reflects on its run and trains the lesson into a LoRA adapter about 0.02% of the model's size. A router later loads the closest adapter when the agent fails a similar task. With Qwen3.6-27B, task success on GAIA rose from 38.8% to 55.8%, beating in-context reflections and a Skills.md file. Across 185 earlier mistakes, 67.3% of repairs still worked by the third encounter, against 42.9% for plain retrying. If your agents keep making the same mistakes, save each fix as a small adapter they can reload, not only as prompt notes.