arXiv cs.CL· Jia Tong, Senyou Deng, Min Peng, Jianfeng Pan·· 3 天前AI 评分26
AtomSearch：基于证据的树搜索优化智能体技能
AtomSearch: Evidence-Grounded Tree Search for Agent Skill Optimization
AI 导读
AtomSearch 通过对比式原子编辑与技能状态树搜索（STS）优化智能体技能，在四个任务和五个目标模型上完成评测。使用 deepseek-v4-flash 时，它在四个任务中的三个上超过 SkillOpt，其中 LiveMathematicianBench 提升 9.68 个百分点，且 SkillOpt 最终技能字符数是其 2.3-19.7 倍。
正文 · 原文
Abstract:Agent skills adapt frozen models through editable external instructions. Revising several rules at once makes changes difficult to inspect, while judging each local edit only by its immediate score can discard useful intermediate states. We introduce AtomSearch, which keeps revision actions local while using subsequent skill states to guide search. Contrastive atomic editing links each primitive rule edit to failed and successful executions, a behavioral hypothesis, and an intended scope. Skill-State Tree Search (STS) evaluates short edit compositions and selectively retains them as executable states that can be revisited and extended. Outcome memory supplies later proposals with observed repair and regression statistics. We evaluate AtomSearch across four tasks and five target models. With deepseek-v4-flash, it exceeds SkillOpt on three of four tasks, including a 9.68-percentage-point gain on LiveMathematicianBench, and SkillOpt's final skills contain 2.3-19.7 times as many characters. In 67% of completed searches, the best validation state passes through an ancestor that was not highest-scoring at its depth. These results show how local rule edits can support the exploration of useful revision paths and yield compact skills, with additional offline evaluations used to explore alternative states.
| Comments: | 15 pages, 5 figures, 5 tables |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.19951 [cs.CL] |
| (or arXiv:2610.19951v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.19951 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jia Tong [view email]
[v1]
Wed, 14 Oct 2026 10:33:30 UTC (244 KB)
来源：arXiv cs.CL · arxiv.org