论文揭示Agent技能的真实作用：稳定执行而非注入知识
Interesting paper demystifying agent skills.
Interesting paper demystifying agent skills.
一篇有趣的论文，揭秘了智能体技能的本质。
If you maintain skills for your agent, this one is worth your time.
如果你为智能体维护技能，这篇值得一读。
(bookmark it)
（建议收藏）
Skills are usually assumed to inject knowledge the model lacks. However, this paper finds something interesting.
通常认为技能是注入模型所缺乏的知识。然而，这篇论文发现了有趣的现象。
Across 8,135 normalized trial records, procedural anchoring accounts for 65.7% of cases where a skill helps, and explicit knowledge injection accounts for 4.5%. Skills stabilize execution rather than supply facts.
在8,135条标准化试验记录中，程序性锚定占技能有效案例的65.7%，而显性知识注入仅占4.5%。技能更多是稳定执行，而非提供事实。
As the pool grows from 5 to 100 skills, actual-use precision falls from 29.6% to 3.3%.
当技能池从5个增至100个时，实际使用精确度从29.6%降至3.3%。
Skills still beat Workflow Memory by 6.06 points in matched comparisons, and they break under brittle assumptions, incompatible contexts, or insufficient adaptation.
在匹配比较中，技能仍比工作流记忆高出6.06分，但在脆弱假设、不兼容上下文或适应不足时，它们会失效。
Paper: https://arxiv.org/abs/2608.14036
论文链接：https://arxiv.org/abs/2608.14036
Track more trending AI papers in our academy: https://academy.dair.ai/
在我们的学院中追踪更多热门AI论文：https://academy.dair.ai/
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力