arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2603.02766cs.AIcs.MA
收藏

EvoSkill:多智能体系统的自动化技能发现

EvoSkill: Automated Skill Discovery for Multi-Agent Systems

Sentient;Virginia Tech(弗吉尼亚理工大学)

Salaheddin Alzubi, Noah Provenzano, Jaydon Bingham, Weiyuan Chen, Tu Vu

AI总结:

EvoSkill通过自进化框架自动发现和优化多智能体系统的技能,提升在OfficeQA和SealQA等基准测试中的性能,并展示技能在跨任务迁移中的有效性。

摘要:

编码代理越来越多地被用作通用问题求解器,但其灵活性本身并不能提供专门任务所需的领域专业知识。最近的工作通过“智能体技能”来解决这个问题:可重用的工作流和代码,这些技能可以增强智能体的领域特定能力。目前大多数技能都是手工编写的,现有的进化方法优化低级制品(例如提示和代码),这些制品与特定模型和任务紧密耦合。我们引入了EvoSkill,一个自进化框架,通过迭代性失败分析自动发现和优化智能体技能。EvoSkill分析执行失败,提出新的技能或对现有技能进行修改,并将它们转化为结构化、可重用的技能文件夹。一个帕累托前沿的智能体程序控制选择,只保留那些在验证性能上有所提升的技能,同时基础模型保持冻结。我们评估了EvoSkill在两个基准测试上的表现:OfficeQA,一个基于美国财政部数据的 grounded reasoning 基准测试,其中它将精确匹配准确率提高了7.3%(60.6%→67.9%);以及SealQA,一个带有噪声检索的搜索增强型问答基准测试,其中它实现了12.1%的提升(26.6%→38.7%)。我们还研究了在一项任务上进化出的技能在另一项任务上的零样本迁移能力;特别是:从SealQA进化出的技能在零样本迁移至BrowseComp时,无需修改就将准确率提高了5.3%,证明了在技能层面的优化能够产生超越训练任务的可迁移能力。

英文摘要:

Coding agents are increasingly used as general-purpose problem solvers, but their flexibility does not by itself confer the domain expertise needed for specialized tasks. Recent work addresses this through \textit{agent skills}: reusable workflows, and code, that augment agents with domain-specific capabilities. Most skills today are hand-crafted, and existing evolutionary approaches optimize low-level artifacts (e.g. prompts \& code) that are tightly coupled to specific models and tasks. We introduce \textbf{EvoSkill}, a self-evolving framework that automatically discovers and refines agent skills through iterative failure analysis. EvoSkill analyzes execution failures, proposes new skills or edits to existing ones, and materializes them into structured, reusable skill folders. A Pareto frontier of agent programs governs selection, retaining only skills that improve held-out validation performance while the underlying model remains frozen. We evaluate EvoSkill on two benchmarks: OfficeQA, a grounded reasoning benchmark over U.S.\ Treasury data, where it improves exact-match accuracy by \textbf{7.3\%} (60.6\% $\to$ 67.9\%); and SealQA, a search-augmented QA benchmark with noisy retrieval, where it yields a \textbf{12.1\%} gain (26.6\% $\to$ 38.7\%). We also investigate the zero-shot transfer capabilties of skills evolved on one task to the other; in particular: skills evolved from SealQA transfers zero-shot to BrowseComp, improving accuracy by \textbf{5.3\%} without modification demonstrating that skill-level optimization produces transferable capabilities beyond the training task.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑