arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2607.01084cs.AI
收藏

智能体能否泛化到开放世界?揭示工具使用中静态训练的脆弱性

Can Agents Generalize to the Open World? Unveiling the Fragility of Static Training in Tool Use

Song-Lin Lv, Weiming Wu, Rui Zhu, Zi-Jian Cheng, Lan-Zhe Guo

AI总结:

针对LLM智能体在真实场景中因查询、工具集和交互动态变化而性能下降的问题,形式化定义了开放世界工具使用问题,构建了四层层次的环境偏移框架,并通过实验揭示了监督微调和强化学习训练智能体在面对开放环境偏移时的脆弱性,提出了扰动增强微调策略以提升鲁棒性。

摘要:

虽然大型语言模型(LLM)智能体在静态基准测试中表现出色,但它们在现实场景中的部署受到用户查询、工具集和交互动态变化的阻碍。为了解决这一泛化差距,我们形式化了OpenAgent(开放世界中的工具使用智能体)问题设置,其特征是查询、动作、观察和领域维度上的分布偏移。为了系统诊断其影响,我们构建了一个受控沙箱环境,在其中定义了跨四个层次层次(感知、交互、推理和内化)的细粒度环境偏移,并进行了全面的实验系列。我们的分析得出一系列关键见解,表明通过监督微调(SFT)和强化学习训练的智能体在面对开放环境偏移时都会遭受不同程度的性能下降。基于这些见解,我们提出了扰动增强微调,一种基于干扰的SFT干预策略,为在现实环境中增强智能体的鲁棒性和实用性奠定了基础。我们的代码将发布在:https://github.com/LAMDA-NeSy/OpenAgent。

英文摘要:

While Large Language Model (LLM) agents demonstrate proficiency in static benchmarks, their deployment in real-world scenarios is hindered by the dynamic nature of user queries, tool sets, and interaction dynamics. To address this generalization gap, we formalize OpenAgent (Tool-Use Agent in Open-World), a problem setting characterized by distributional shifts across query, action, observation, and domain dimensions. To systematically diagnose its impact, we construct a controlled sandbox environment where we define fine-grained environmental shifts across a four-tier hierarchy, Perception, Interaction, Reasoning, and Internalization, and conduct a comprehensive series of experiments. Our analysis yields a series of key insights, demonstrating that agents trained via both Supervised Fine-Tuning(SFT) and Reinforcement Learning suffer from varying degrees of performance degradation when confronting open environmental shifts. Building on these insights, we propose Perturbation-Augmented Fine-Tuning, a disturbance-based intervention strategy for SFT that lays the foundation for enhancing agent robustness and utility in realistic environments. Our code will be released at: https://github. com/LAMDA-NeSy/OpenAgent.

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。

↑