发现大模型智能体技能库的缩放规律，提升任务调度与执行效率。
The Scaling Laws of Skills in LLM Agent Systems
- 提出路由与执行双规律，揭示技能库规模对准确率的影响机制。
- 优化后路由准确率从71.3%提升至91.7%，劫持率由22.4%降至4.1%。
- 适用于复杂任务系统设计，尤其适合构建可复用的智能体技能库。
随着智能体系统规模扩大，技能逐渐积累成大型可复用库，但其缩放规律尚不明确。在15个前沿大模型、1,141项真实世界技能及超300万次路由或执行决策中，我们识别出两条耦合规律：路由规律显示，单步路由准确率随技能库规模呈对数衰减（所有模型的R² > 0.97），错误类型从局部技能竞争演变为跨家族漂移，并被过于通用的“黑洞技能”捕获；执行规律表明，状态未实现前，联合路由近似乘积关系，而正确执行可使困难下游任务决策性能提升约4倍。一个参数——路由对数衰减斜率b——将两者耦合：路由侧拟合可预测执行侧恢复能力，说明同一库属性同时控制预执行崩溃与下游可恢复性。该规律具有可操作性：基于规律优化后，保留测试集路由准确率从71.3%升至91.7%，劫持率从22.4%降至4.1%，并在下游ClawBench和ClawMark任务中定向迁移，平均通过率分别从49.3%提升至61.6%，28.4%提升至34.5%。结果表明，智能体性能不仅取决于模型能力，更受技能库结构、粒度与暴露策略影响。
原文摘要 · Abstract (English)
As agent systems scale, skills accumulate into large reusable libraries, yet their scaling laws remain poorly understood. Across 15 frontier LLMs, 1,141 real-world skills, and over 3M routing or execution decisions, we identify two coupled laws. Routing law: single-step routing accuracy decays logarithmically with library size ($R^2{>}0.97$ for all models), with errors progressing from local skill competition to cross-family drift and capture by overly general "black-hole skills". Execution law: before state realization, joint routing is approximately multiplicative, whereas correct execution can improve difficult downstream decisions by about $4{\times}$. A single parameter, the routing logarithmic decay slope $b$, couples the two laws: routing-side fits predict execution-side rescue across models, showing that the same library property controls both pre-execution collapse and downstream recoverability. The laws are actionable: law-guided optimization raises held-out routing accuracy from 71.3% to 91.7%, reduces hijack from 22.4% to 4.1%, and transfers directionally to downstream ClawBench and ClawMark execution settings, improving mean pass rate from 49.3% to 61.6% on ClawBench and from 28.4% to 34.5% on ClawMark. These results show that agent performance depends not only on model capability, but also on the structure, granularity, and exposure policy of the skill library.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。