arXiv cs.CL· Jiahao Wang, Yuqi Xiong, Qing Yang, Zhipeng Xu, Zhenghao Liu, Zulong Chen, Yubin Bao·· 10 小时前AI 评分31
PASE：程序辅助的可补丁技能进化框架
Program-Aided Patchable Skill Evolution
AI 导读
研究者提出 PASE（Program-Aided Patchable Skill Evolution），将 LLM 智能体技能中的可复用流程逻辑拆分为可编辑程序，从而对可执行组件做定向更新，减少反复修改自然语言技能文档。在多种模型和基准上，PASE 一致优于所有基线，提升超过 2.8%，并表现出更好的进化稳定性与泛化能力。全部数据与代码将通过 GitHub 发布。
正文 · 原文
Abstract:Agent skills have emerged as an effective means of enabling LLM agents to solve complex, interactive tasks by encoding reusable knowledge without modifying model parameters. Recent approaches further improve skill adaptation by distilling reusable experience from agent execution trajectories and evolving skills based on downstream task feedback. However, existing methods primarily represent skills as natural language documents, intertwining procedural logic with descriptive knowledge and making skills difficult to maintain and evolve. To address this limitation, we propose PASE (Program-Aided Patchable Skill Evolution), a program-aided framework that iteratively evolves skills by associating them with executable programs. Unlike prior approaches that encode trajectory-derived experience primarily in natural language, PASE separates reusable procedural logic into editable programs, enabling targeted updates to executable components while reducing the need to repeatedly revise natural language skills. Extensive experiments across diverse models and benchmarks demonstrate that PASE consistently outperforms all baselines, by achieving over 2.8% improvements. Further analyses show that PASE exhibits greater evolutionary stability and generalisation capability, highlighting the critical role of programmatic knowledge in skill evolution. We will release all data and code via GitHub.
| Comments: | 17 pages, 6 figures, 5 tables |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.14019 [cs.CL] |
| (or arXiv:2611.14019v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.14019 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Qing Yang [view email]
[v1]
Tue, 10 Nov 2026 12:53:12 UTC (800 KB)
来源：arXiv cs.CL · arxiv.org