arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.00851cs.AIcs.MA
收藏

理解长期运行代理中的说服

Persuasion Propagation: Does Persuasion Change AI Agent Behavior?

University of Massachusetts Amherst(马萨诸塞大学阿默斯特分校)

Hyejun Jeong, Amir Houmansadr, Shlomo Zilberstein, Eugene Bagdasarian

AI总结:

本文研究了长期任务中代理受到用户说服影响的行为变化,提出了一种基于行为的评估框架,发现提前指定信念状态的代理在搜索和源访问上表现更高效,表明说服影响代理行为。

摘要:

现代AI代理越来越多地结合对话交互与自主任务执行,例如编码和网络研究,这引发了一个自然问题:当一个从事长期任务的代理受到用户说服时会发生什么?然而研究这一可能性具有挑战性,因为长期运行的代理行为具有噪声且难以重复,而且不清楚只有在扩展任务执行中才会出现哪些独特挑战。我们研究了信念层面干预如何影响下游任务行为,这种现象我们称之为说服传播。我们介绍了一种以行为为中心的评估框架,区分在任务执行期间或之前应用的说服。在网页研究和编码任务中,我们发现即时说服导致的行为影响弱且不一致。相反,当在任务时间显式指定信念状态时,信念预填充的代理平均进行26.9%更少的搜索,并访问16.9%更少的唯一来源,比中性预填充的代理。这些结果表明,即使在之前的交互中,说服也会影响代理的行为,从而推动对代理系统的行为层面评估。

英文摘要:

AI agents combine normal conversation with autonomous task execution, allowing earlier context to shape how later tasks are executed. Yet studying this possibility is challenging because agent behavior is noisy and costly to reproduce, and observed behavioral changes can be difficult to separate from generic context sensitivity. We first ask whether task-irrelevant persuasion changes how agents act, a phenomenon we call \emph{persuasion propagation}, and whether persuasion susceptibility explains variation in these effects. We show that persuasive interaction leaves a measurable, task-dependent behavioral footprint on downstream agent execution. In web research, it increases execution time by 56\% and domains visited by 16.3\% relative to a baseline, while coding shows faster execution with larger revisions. However, agents that are more susceptible to persuasive conversational context do not reliably predict the magnitude of the downstream behavioral shifts. This motivates behavior-level evaluation of persuasion in AI agents.

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。

↑