arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2605.26302cs.AIcs.CLcs.MA
收藏

你的智能体也在老化:面向部署系统的智能体寿命工程

Your Agents Are Aging Too: Agent Lifespan Engineering for Deployed Systems

The University of Texas at Austin(德克萨斯大学奥斯汀分校)

Jianing Zhu, Yeonju Ro, John Robertson, Kevin Wang, Junbo Li, Haris Vikalo, Aditya Akella, Zhangyang Wang

AI总结:

提出 AgingBench 基准,通过四种老化机制和诊断工具评估部署后智能体的可靠性退化,并指出需要寿命评估、机制级诊断和阶段针对性修复。

摘要:

长寿命AI智能体越来越多地被部署为持久化运行系统,但它们仍然像刚初始化的模型一样被评估。第一天基准测试忽略了一个基本系统问题:智能体在部署后能保持可靠多久?即使模型权重被冻结,智能体的有效状态也在不断变化,因为它压缩交互历史、从不断增长的记忆库中检索、在更新后修正事实,并经历常规维护。因此,可靠性成为整个智能体框架的寿命属性,而不仅仅是基础模型的快照属性。我们引入了AgingBench,一个用于智能体寿命工程的纵向可靠性基准:不仅测量部署的智能体是否退化,还测量退化的形式以及修复应针对何处。AgingBench将智能体老化组织为四种机制:压缩老化、干扰老化、修订老化和维护老化。为了诊断这些故障,AgingBench使用时间依赖图和对偶反事实探针,为记忆管道的写入、检索和利用阶段生成诊断档案。在7个场景、14个模型、多种记忆策略以及运行者控制和自主智能体上,跨越约400次运行(涵盖8到200个会话)的结果表明,智能体老化不是一维的:行为测试可以保持干净,而事实精度下降;派生状态跟踪可能在单个模型内急剧崩溃;相同的错误答案可能需要不同的修复,具体取决于诊断档案指向的内容。这些结果表明,可靠的智能体部署需要寿命评估、机制级诊断和阶段针对性修复,而不仅仅是更强的第一天模型。

英文摘要:

Long-lived AI agents are increasingly deployed as persistent operational systems, yet they are still mostly evaluated like freshly initialized models. Day-one benchmarks miss a basic systems question: how does an agent's reliability change after deployment? Even when model weights are frozen, an agent's effective state keeps changing as it compresses interaction history, retrieves from a growing memory store, revises facts after updates, and undergoes routine maintenance. Reliability therefore becomes a lifespan property of the full agent harness, not only a snapshot property of the base model. We introduce AgingBench, a longitudinal reliability benchmark for agent lifespan engineering: measuring not only whether deployed agents degrade, but what form the degradation takes and where repair should target. AgingBench organizes agent aging into four mechanisms: compression aging, where writing and rewriting drops future-relevant details; interference aging, where accumulated similar memories crowd out the target fact; revision aging, where changed or derived state is not updated correctly; and maintenance aging, where lifecycle events such as model swapping trigger regressions. To diagnose these failures, AgingBench uses temporal dependency graphs and counterfactual probes that produce diagnostic profiles for the write, retrieval, and utilization stages of the memory pipeline. Across 6 scenarios, 10 memory systems and 6 models from two families, we find that agent aging is not one-dimensional over 400 runs spanning 50-200 sessions: a well-formed reply can hide a value the memory no longer supplies; derived state collapses as updates accumulate; and the same wrong answer can require different repairs depending on what the diagnostic profile points to.

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。

↑