Expired65% confidenceFactExact time
When available tools exceed 15-20, mainstream LLMs show a noticeable decline in tool selection accuracy.
3
Sources
65%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026(expired)
Sources
Core Analysis of Agent Skill: Progressive Disclosure & Middleware Practical Guide
bilibili码士集团-马小雪6/22/2026
Related Claims
Unverified在工具数量超过20个的场景下，工具描述的语义歧义度与模型选择准确率之间存在显著的负相关关系73% similarUnverifiedIn real-environment A/B testing, the post-trained checkpoint reduced the tool-call failure rate by 21.2% relatively compared to the earlier version.67% similarUnverifiedThe Gorilla benchmark released by the University of California, Berkeley shows that when the number of candidate tools exceeds 20, even models at the GPT-4 level experience a 15%-30% drop in tool selection accuracy.65% similarUnverifiedOn the rare entity slices newly partitioned by structural metrics, the accuracy of current state-of-the-art systems drops by 15.4% to 39.9%.63% similarUnverified模型自主触发Skills时，全量注入所有Skill描述的方案在Skills数量超过数十个后性能显著下降63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/42523API
curl https://kongchang.com/api/v1/knowledge/claims/42523MCP
get_claim(id=42523)