AtomSearch：用证据驱动的树搜索优化 Agent 技能
AtomSearch: Evidence-Grounded Tree Search for Agent Skill Optimization
一篇讲怎么高效改 Agent 技能规则的论文，AtomSearch 在 deepseek-v4-flash 上比 SkillOpt 高 9.68 个百分点，技能还短好几倍。
AtomSearch 针对冻结模型的技能优化问题，通过对比式原子编辑把每条规则修改与失败/成功执行、行为假设和作用范围绑定。其 Skill-State Tree Search 会评估短编辑组合，保留可复用、可扩展的中间状态，再用结果记忆为后续提议提供修复与回退统计。在 4 个任务、5 个目标模型的评测中，deepseek-v4-flash 上有 3 个任务超过 SkillOpt，其中 LiveMathematicianBench 提升 9.68 个百分点，且 SkillOpt 最终技能文本长度是它的 2.3-19.7 倍。67% 的搜索中最优验证状态经过的非最高分祖先节点证明局部编辑路径有探索价值。
AtomSearch: Evidence-Grounded Tree Search for Agent Skill Optimization
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.
