[论文] TimeEvo: Failure-Driven Self-Evolution of a Time Series Agent

论文概要

研究领域: ML 作者: Jie Yang, Yan Zheng, Jiarui Sun, Xiran Fan, Junpeng Wang, Liang Wang, Zelin Xu, Qinghua Liu, Zhengyu Fang, Yiwei Cai, Philip S. Yu 发布时间: 2026-09-25 arXiv: 2609.27277

中文摘要

时间序列智能体通过调用外部工具回答分析问题，而携带哪些工具由人们在智能体运行前决定。然而我们识别出该设定中的两个失效。人机工具错配：含 21 个专家策划工具的库在一些任务上有帮助、在另一些任务上反而有害，在我们测试的每个骨干网络上都压低了异常检测准确率。静默伤害：一轮通用自我修订改变了 147 个答案并改坏其中 56 个，而最终分数变动不到一分。两者源于同一缺口：一个工具有用与否是运行时逐题决定的，而工具是预先提供的、仅凭单一平均值被评判。为此我们提出 TimeEvo：将智能体的已诊断故障聚类为能力缺口，为每个缺口规划一项测量，合成仅含证据的工具填补缺口，并仅通过一道成对准入门来接纳候选工具库。十个时间序列 QA 任务、三个骨干网络上的实验表明：TimeEvo 从空库起步，在每个任务、每个骨干上都提升了准确率；在廉价模型上成长的工具库装入更强模型后仍有收益。代码开源于论文中的 https URL。

原文摘要

Time series agents answer analytical questions by calling external tools, and which tools they carry is decided by people before the agent runs. However, we identify two failures in this setup. Human-Agent Tool Misalignment: a library of 21 expert-curated tools helps on some tasks and hurts on others, dropping anomaly accuracy under every backbone we test. Silent Harm: one round of generic self-revision changes 147 answers and breaks 56 of them, while the final score moves by less than a point. Both follow from the same gap: whether a tool helps is decided question by question at runtime, while tools are supplied in advance and judged by a single average. To address this, we propose TimeEvo, which clusters an agent's diagnosed failures into capability gaps, plans a measurement for each...

#论文 #arXiv #ML #小凯