LLM智能体技能编排可编译多少？基于执行的特化实现高效token技能使用
原标题：How Much Skill Orchestration Can LLM Agents Compile Away? Execution-Grounded Specialization for Token-Efficient Skill Use
AI 摘要
该论文研究LLM智能体能否将技能编排自动编译为可执行快速路径。作者发现零样本技能特化因运行时绑定假设错误而效果不佳，提出SkillCache方法，利用单次执行轨迹将技能编译为参数化胶囊。在Claude Code环境下，87个SkillsBench任务中重复请求的token消耗降低81%（Sonnet 4.6）和74%（Opus 4.8），且胶囊可跨模型迁移，使三个开源模型平均奖励提升12-25个百分点、token减少72-81%。
正文节选
marginparsep has been altered. topmargin has been altered. marginparwidth has been altered. marginparpush has been altered. The page layout violates the style. Please do not change the page layout, or include packages like geometry, savetrees, or fullpage, which change it for you. We’re not able to reliably undo arbitrary changes to the style. Please remove the offending package(s), or layout-changing commands and try again. How Much Skill Orchestration Can LLM Agents Compile Away? Execution-Gro