When Errors Become Narratives: A Longitudinal Taxonomy of Silent Failures in a Production LLM Agent Runtime
当错误成为叙事:生产级LLM Agent运行时中静默故障的纵向分类
机构 * Independent researcher(独立研究者)
专题命中 软件智能体 :agent(title,title_cn);分类 cs.AI、cs.SE
AI总结 通过八周对生产级个人助手Agent运行时的研究,识别出28次静默故障,提出五类机制导向分类,其中D类(链式幻觉与捏造)为LLM特有且最危险,系统会生成流畅可信的虚假叙事。
Comments 18 pages, 5 figures, 2 tables. 22 incident postmortems and all defense-framework artifacts publicly available at https://github.com/bisdom-cell/openclaw-model-bridge; governance engine on PyPI (openclaw-ontology-engine)