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 2510.01670cs.AIcs.CLcs.CRcs.CYcs.LG
收藏

只管去做!?计算机使用智能体展现出盲目目标导向性

Just Do It!? Computer-Use Agents Exhibit Blind Goal-Directedness

Microsoft Research AI Frontiers(微软研究院人工智能前沿);Microsoft AI Red Team(微软AI红色团队);University of California, Riverside(加州大学河滨分校);NVIDIA(英伟达)

Erfan Shayegani, Keegan Hines, Yue Dong, Nael Abu-Ghazaleh, Roman Lutz, Spencer Whitehead, Vidhisha Balachandran, Besmira Nushi, Vibhav Vineet

AI总结:

本研究发现计算机使用智能体普遍存在盲目目标导向性偏差,归纳三类典型模式,构建BLIND-ACT基准评估9个前沿模型发现平均偏差率达80.8%,揭示潜在安全风险与失效模式,为风险缓解研究奠定基础。

摘要:

计算机使用智能体(Computer-Use Agents, CUAs)是一类应用日益广泛的智能体,它们通过在图形用户界面(GUI)上执行操作来完成用户目标。本文研究表明,CUAs 普遍表现出盲目目标导向性(Blind Goal-Directedness, BGD):即不顾可行性、安全性、可靠性或上下文语境,一味追求目标的偏差行为。我们归纳了 BGD 的三种常见模式:(i)缺乏上下文推理能力,(ii)在模糊情境下做出假设和决策,(iii)目标存在矛盾或不可行。我们构建了 BLIND-ACT 基准,包含90项任务,覆盖上述三种模式。该基准基于 OSWorld 搭建,提供真实的交互环境,并采用基于大语言模型(LLM)的评判器评估智能体行为,与人工标注的一致性达到93.75%。我们使用 BLIND-ACT 评估了9个前沿模型,包括 Claude Sonnet、Opus 4、Computer-Use-Preview 和 GPT-5,观测到这些模型的平均 BGD 发生率高达80.8%。研究表明,即使输入内容不具有直接危害性,BGD 也会引发潜在风险。尽管基于提示词的干预手段能降低 BGD 水平,但仍存在大量风险,凸显了强化训练阶段或推理阶段干预措施的必要性。定性分析揭示了三类已观测到的失效模式:执行优先偏差(更关注如何行动而非是否该行动)、思行脱节(执行过程与推理结果相悖)、请求至上(以用户请求为行动正当理由)。对 BGD 的识别与 BLIND-ACT 基准的提出,为未来研究和缓解这一根本性风险、保障 CUA 安全部署奠定了基础。

英文摘要:

Computer-Use Agents (CUAs) are an increasingly deployed class of agents that take actions on GUIs to accomplish user goals. In this paper, we show that CUAs consistently exhibit Blind Goal-Directedness (BGD): a bias to pursue goals regardless of feasibility, safety, reliability, or context. We characterize three prevalent patterns of BGD: (i) lack of contextual reasoning, (ii) assumptions and decisions under ambiguity, and (iii) contradictory or infeasible goals. We develop BLIND-ACT, a benchmark of 90 tasks capturing these three patterns. Built on OSWorld, BLIND-ACT provides realistic environments and employs LLM-based judges to evaluate agent behavior, achieving 93.75% agreement with human annotations. We use BLIND-ACT to evaluate nine frontier models, including Claude Sonnet and Opus 4, Computer-Use-Preview, and GPT-5, observing high average BGD rates (80.8%) across them. We show that BGD exposes subtle risks that arise even when inputs are not directly harmful. While prompting-based interventions lower BGD levels, substantial risk persists, highlighting the need for stronger training- or inference-time interventions. Qualitative analysis reveals observed failure modes: execution-first bias (focusing on how to act over whether to act), thought-action disconnect (execution diverging from reasoning), and request-primacy (justifying actions due to user request). Identifying BGD and introducing BLIND-ACT establishes a foundation for future research on studying and mitigating this fundamental risk and ensuring safe CUA deployment.

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。

↑