arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.02760cs.CLcs.LG
收藏

从任务解决到在LLM代理中的鲁棒现实适应

From Task Solving to Robust Real-World Adaptation in LLM Agents

Megagon Labs(梅加戈恩实验室)

Pouya Pezeshkpour, Estevam Hruschka

AI总结:

本文研究了LLM代理在现实环境中的鲁棒适应能力,发现任务解决与实际部署鲁棒性存在显著差距,揭示了在部分可观测性和噪声环境下代理的决策挑战。

摘要:

大型语言模型越来越多地被部署为专门的代理,这些代理能够规划、调用工具并执行长时间跨度的操作。然而,许多现有的评估假设了一个“干净界面”,其中动态被指定且稳定,工具和传感器是可靠的,成功由单一明确的目标捕捉——这往往高估了现实世界的准备程度。在实践中,代理面临规则不明确、信号不可靠、环境变化以及隐含的多利益相关者目标。因此,挑战不仅仅是解决任务,而是在解决过程中进行适应:决定信任什么、想要什么、何时验证、何时退后或升级。我们强调测试在四个操作情况下部署相关的鲁棒性:部分可观测性、动态环境、噪声信号和动态代理状态。我们在一个基于网格的游戏中对代理LLM进行基准测试,该游戏具有简单的目标但长时间跨度的执行。剧集违反了干净界面的假设,但仍然可解决,迫使代理推断规则、支付信息费用、适应环境和内部变化,并在噪声下谨慎行动。在五个最先进的LLM代理中,我们发现名义任务解决与部署类似鲁棒性之间存在显著差距。表现通常随着网格大小和时间跨度的增加而下降,但排名不稳定:较弱的模型在策略匹配不确定性领域时可以击败更强的模型。尽管没有明确的指令,代理在完成、效率和惩罚避免之间进行权衡,表明部分目标推断。消融和特征分析揭示了模型特定的敏感性和失败驱动因素,推动了在部分可观测性、噪声和非平稳性下进行验证、安全行动选择和目标推断的工作。

英文摘要:

Large language models are increasingly deployed as specialized agents that plan, call tools, and take actions over extended horizons. Yet many existing evaluations assume a "clean interface" where dynamics are specified and stable, tools and sensors are reliable, and success is captured by a single explicit objective-often overestimating real-world readiness. In practice, agents face underspecified rules, unreliable signals, shifting environments, and implicit, multi-stakeholder goals. The challenge is therefore not just solving tasks, but adapting while solving: deciding what to trust, what is wanted, when to verify, and when to fall back or escalate. We stress-test deployment-relevant robustness under four operational circumstances: partial observability, dynamic environments, noisy signals, and dynamic agent state. We benchmark agentic LLMs in a grid-based game with a simple goal but long-horizon execution. Episodes violate clean-interface assumptions yet remain solvable, forcing agents to infer rules, pay for information, adapt to environmental and internal shifts, and act cautiously under noise. Across five state-of-the-art LLM agents, we find large gaps between nominal task-solving and deployment-like robustness. Performance generally degrades as grid size and horizon increase, but rankings are unstable: weaker models can beat stronger ones when strategy matches the uncertainty regime. Despite no explicit instruction, agents trade off completion, efficiency, and penalty avoidance, suggesting partial objective inference. Ablations and feature analyses reveal model-specific sensitivities and failure drivers, motivating work on verification, safe action selection, and objective inference under partial observability, noise, and non-stationarity.

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。

↑