arXiv cs.AI· Jinlin Liu, Wei Chen, Xiaojin Zhang·· 3 天前AI 评分44
LLM 智能体为何在技能库扩大时无法加载相关技能：隐式路由的失效
Lost in Implicit Routing: How LLM Agents Fail to Load Relevant Skills at Scale
AI 导读
研究发现 LLM 智能体的技能路由崩溃：在 SkillsBench 任务和黄金技能固定不变的情况下加入无关技能，加载技能召回率从 73.6% 降至 27.4%，零加载率从 9.3% 升至 51.6%，而显式预选仍保持 73.9% 召回率。仅遮蔽无关技能的描述可基本恢复技能加载，表明存在描述层面的语义干扰与技能选择同任务执行之间的指令干扰。
正文 · 原文
Abstract:Reusable skills let LLM agents load task-specific instructions on demand, but they make execution depend on implicit routing over skill metadata. We test whether this design scales by holding SkillsBench tasks and their golden skills fixed while adding irrelevant skills. We find skill routing collapse: agents increasingly load fewer or no skills, even though the relevant skills remain available. With 1,000 irrelevant skills, loaded-skill recall falls from 73.6% to 27.4% and the zero-load rate rises from 9.3% to 51.6%, whereas explicit selection from the same library retains 73.9% recall. Loading other skills, including plausible alternatives to golden skills, is too rare to explain this decline. Masking only the descriptions of irrelevant skills largely restores skill loading, yet with the metadata intact, explicitly asking agents to plan skill use still yields lower recall than preselection. These findings point to semantic interference from descriptions and instruction interference when skill choice competes with task execution. We also evaluate explicit preselection as a practical mitigation, including its coverage and latency trade-offs. Our analysis informs skill protocols and evaluation for agents with growing skill libraries.
| Comments: | 22 pages, including references and appendices. Code: this https URL |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.18522 [cs.AI] |
| (or arXiv:2610.18522v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.18522 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jinlin Liu [view email]
[v1]
Tue, 13 Oct 2026 15:12:12 UTC (1,765 KB)
来源：arXiv cs.AI · arxiv.org