arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2603.03378cs.LGcs.AI
收藏

AOI: 将失败轨迹转化为自主云诊断的训练信号

AOI: Turning Failed Trajectories into Training Signals for Autonomous Cloud Diagnosis

Gradient;Soochow University(苏州大学);UC Santa Cruz;Georgia Institute of Technology(佐治亚理工学院);University College London(伦敦大学学院);WeJoy;ByteDance(字节跳动)
Pei Yang, Wanyi Chen, Asuka Yuxi Zheng, Xueqian Li, Xiang Li, Haoqin Tu, Jie Xiao, Yifan Pang, Dongdong Zhang, Fuqiang Li, Alfred Long, Lynn Ai, Eric Yang, Bill… 展开作者

Pei Yang, Wanyi Chen, Asuka Yuxi Zheng, Xueqian Li, Xiang Li, Haoqin Tu, Jie Xiao, Yifan Pang, Dongdong Zhang, Fuqiang Li, Alfred Long, Lynn Ai, Eric Yang, Bill Shi, Yifan Chen, Shuhan Sun

AI总结:

AOI通过结构化轨迹学习框架,在安全约束下实现自动化操作,结合可训练诊断系统、读写分离执行架构和失败轨迹闭环演进器,提升云诊断性能。

摘要:

大型语言模型(LLM)代理提供了一种有前途的数据驱动方法来自动化站点可靠性工程(SRE),但其企业部署受限于三个挑战:对专有数据的访问受限、受权限管理环境下的不安全行动执行,以及封闭系统无法从失败中改进。我们提出了AOI(自主操作智能),一种可训练的多智能体框架,将自动化操作建模为在安全约束下的结构化轨迹学习问题。我们的方法整合了三个关键组件。首先,可训练的诊断系统应用群相对策略优化(GRPO)将专家级知识蒸馏到本地部署的开源模型中,使基于偏好的学习无需暴露敏感数据。其次,读写分离的执行架构将操作轨迹分解为观察、推理和行动阶段,允许安全学习并防止未经授权的状态突变。第三,失败轨迹闭环演进器挖掘不成功的轨迹并将其转化为纠正监督信号,实现持续的数据增强。在AIOpsLab基准测试中,我们的贡献实现了累积收益。(1)AOI运行时单独在所有86个任务上实现了66.3%的best@5成功,优于先前的最先进水平(41.9%)24.4分。 (2)添加观察者GRPO训练,本地部署的14B模型在63个未见故障类型的保留任务上实现了42.9%的avg@1,优于Claude Sonnet 4.5。 (3)演进器将37条失败轨迹转化为诊断指导,使端到端avg@5提高4.8分,同时减少方差35%。

英文摘要:

Large language model (LLM) agents offer a promising data-driven approach to automating Site Reliability Engineering (SRE), yet their enterprise deployment is constrained by three challenges: restricted access to proprietary data, unsafe action execution under permission-governed environments, and the inability of closed systems to improve from failures. We present AOI (Autonomous Operations Intelligence), a trainable multi-agent framework formulating automated operations as a structured trajectory learning problem under security constraints. Our approach integrates three key components. First, a trainable diagnostic system applies Group Relative Policy Optimization (GRPO) to distill expert-level knowledge into locally deployed open-source models, enabling preference-based learning without exposing sensitive data. Second, a read-write separated execution architecture decomposes operational trajectories into observation, reasoning, and action phases, allowing safe learning while preventing unauthorized state mutation. Third, a Failure Trajectory Closed-Loop Evolver mines unsuccessful trajectories and converts them into corrective supervision signals, enabling continual data augmentation. Evaluated on the AIOpsLab benchmark, our contributions yield cumulative gains. (1) The AOI runtime alone achieves 66.3% best@5 success on all 86 tasks, outperforming the prior state-of-the-art (41.9%) by 24.4 points. (2) Adding Observer GRPO training, a locally deployed 14B model reaches 42.9% avg@1 on 63 held-out tasks with unseen fault types, surpassing Claude Sonnet 4.5. (3) The Evolver converts 37 failed trajectories into diagnostic guidance, improving end-to-end avg@5 by 4.8 points while reducing variance by 35%.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑