arXiv cs.MA· Aiden Rubey, Zachary Coalson, Cong Chen, Jean-Philippe Monteuuis, Jonathan Petit, Sanghyun Hong·· 6 天前AI 评分33
硬件故障在 LLM 智能体中如何呈现不同表现？
How Do Hardware Faults Manifest Differently in LLM Agents?
AI 导读
首个针对 LLM 智能体硬件故障韧性的系统性研究，基于三个智能体 benchmark、6,814 次执行（2,400+ 次无故障运行与 4,300+ 次注入随机双比特权重故障的运行）追踪故障传播路径。研究发现智能体执行会把模型级损坏转化为系统级结果：智能体可从故障输出中恢复并正确完成任务，但计算与执行成本上升；少数故障会引发语义上微妙但有意义的错误，且在某些模型和场景下会改变安全与对齐行为。
正文 · 原文
Abstract:We present the first systematic study of the resilience of large language model (LLM) agents to random hardware-induced faults. Prior work on hardware fault resilience in LLMs focuses on model-level outcomes, such as incorrect outputs or generation failures. Yet LLM agents repeatedly interact with tools and environments, creating loops that can change how hardware faults manifest. We conduct an extensive evaluation across three agent benchmarks, comprising 6,814 agent executions: 2,400+ clean runs and 4,300+ fault-injected runs with independently sampled random two-bit weight faults. We trace how these faults propagate through agent execution. Unlike prior studies, we find that agentic execution transforms model-level corruption into distinct system-level outcomes: (1) agents can recover from faulty model outputs and still complete tasks correctly; (2) this recovery increases computational and execution costs; (3) a few faults cause subtle but semantically meaningful errors while the agent otherwise behaves normally; and (4) for some models and scenarios, faults can alter security and alignment behavior.
| Comments: | Accepted to NeurIPS 2026 MLforSystems and AI4Good Workshops; The first two authors contributed equally |
| Subjects: | Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2610.12756 [cs.MA] |
| (or arXiv:2610.12756v1 [cs.MA] for this version) | |
| https://doi.org/10.48550/arXiv.2610.12756 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Sanghyun Hong [view email]
[v1]
Thu, 8 Oct 2026 22:08:08 UTC (198 KB)
来源：arXiv cs.MA · arxiv.org