arXiv cs.AI· \"Omer Faruk Akg\"ul, Shuli Jiang, Yuting Zhang, Yi Zhang, Xiaohan Fei, Wei Xia, Stefano Soatto·· 4 小时前AI 评分39
当 Agent 技能过时:动态环境下的运行时行为与选择性维护
When Agent Skills Go Stale: Runtime Behavior and Selective Maintenance in Evolving Environments
AI 导读
环境变化后,原本有效的 Agent 技能可能使任务性能下降超 10 个百分点,或让执行成本增加近 40%。研究提出选择性维护方法,无需额外模型调用即可重放成功执行中的检查,验证观测差异是否与技能状态冲突并修复受影响内容,相比复用陈旧状态最多提升任务准确率 67pp、减少执行步骤 37%。
正文 · 原文
Abstract:Agent skills use past experience and current observations to guide decisions and actions on related tasks. This experience is retained in skill state, whose reuse can amortize inference-time reasoning and environment interaction. When environment changes, parts of this state can become outdated while others remain useful. Across four task settings and two models, we compare skill reuse before and after environment changes. Previously beneficial skills can reduce task performance by over 10 percentage points or increase execution cost by nearly 40% relative to no skill. Task success can also remain unchanged while execution savings decline. These effects depend on which environment-specific details become outdated and the agent's ability to compensate using current observations. We introduce selective maintenance, which replays checks from successful executions without additional model calls. It verifies whether observed differences conflict with skill state, repairs affected content, and validates updates, rebuilding when local repair is insufficient. No description of the change is supplied. Relative to reusing stale state, maintenance improves task accuracy by up to 67pp and reduces execution steps by up to 37%. These results suggest that preserving a skill's inference time value as environments evolve requires selectively maintaining its state.
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.22525 [cs.AI] |
| (or arXiv:2611.22525v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.22525 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Ömer Faruk Akgül [view email]
[v1]
Tue, 17 Nov 2026 10:18:46 UTC (656 KB)
来源:arXiv cs.AI · arxiv.org