arXiv：cs.AI· Jiale Liu, Huajun Xi, Shaokun Zhang, Yifan Zeng, Tianwei Yue, Chi Wang, Jian Kang, Qingyun Wu, Huazheng Wang·· 13 小时前
Who&When Pro：LLM 真能定位 AI 智能体的失败吗？
Who&When Pro: Can LLMs Really Attribute Failures in AI Agents?
AI 导读
研究者提出 Who&When Pro，一个面向 AI 智能体自动失败归因的大规模基准，包含 12,326 条失败轨迹和 golden labels，覆盖 3 种模态、26 个基准。该基准通过严格控制的流程构建：只有在完整重放成功前缀后才注入失败。实验还分析了模型在不同模态、协议和模型族间的归因模式。
正文
Abstract:Automated failure attribution uses LLMs to identify where and why agentic systems fail. As agents become more capable, their failures become subtler, making automated attribution increasingly important. We introduce Who&When Pro, a large-scale benchmark for automated failure attribution in agentic systems. Using a strictly controlled pipeline that injects a failure only after exactly replaying a successful prefix, we construct 12,326 failed trajectories with golden labels across 3 modalities and 26 benchmarks covering various scenarios. Beyond benchmarking, we conduct extensive experiments and analyses, revealing systematic patterns in how models attribute failures across modalities, protocols, and model families, and providing empirical guidance for future automated failure attribution systems.
| Subjects: | Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2607.09996 [cs.AI] |
| (or arXiv:2607.09996v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2607.09996 arXiv-issued DOI via DataCite |
Submission history
From: Jiale Liu [view email]
[v1]
Fri, 10 Jul 2026 21:45:16 UTC (2,480 KB)
[v2]
Fri, 9 Oct 2026 04:04:24 UTC (2,462 KB)
来源：arXiv：cs.AI · arxiv.org