GRUET: Quantifying Uncertainty of Agentic Reasoning-and-Acting Processes
作者: Shuang Liang, Xin-Yu Hu, Shao-Qun Zhang
分类: cs.AI
发布日期: 2026-09-21
💡 一句话要点
提出GRUET以量化智能体推理与行动过程的不确定性
🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 智能体 推理与行动 不确定性量化 大型语言模型 图模型 轨迹分析 性能评估
📋 核心要点
- 现有方法在智能体推理与行动过程中面临显著的不确定性,导致相同任务产生不同轨迹,影响智能体的可信度。
- 本文提出GRUET方法,通过建模推理空间为图,量化轮次级推理不确定性,并聚合轨迹级不确定性。
- 实验证明,GRUET在选择性生成性能上优于现有方法,使用AUROC、AUPRC和AUARC等指标进行评估。
📝 摘要(中文)
智能体因其在开放和动态环境中执行推理与行动(ReAct)的能力而受到广泛关注。ReAct过程通常表现为多轮轨迹,其中大型语言模型(LLMs)生成推理链和任务特定的行动。然而,智能体常常面临显著的不确定性,相同任务可能产生不同的轨迹;高不确定性的轨迹往往导致不可理解的行为,严重影响智能体的可信度。本文推测,这种轨迹级不确定性通常源于LLMs引起的累积轮次级推理不确定性。基于此,我们提出了基于图的轨迹推理不确定性量化方法(GRUET),包括轮次级推理不确定性量化和轨迹级不确定性聚合。通过对九个LLMs和五个基准的实证评估,验证了GRUET在选择性生成性能上的有效性。
🔬 方法详解
问题定义:本文旨在解决智能体推理与行动过程中的不确定性问题,现有方法未能有效量化推理和行动的多轮轨迹中的不确定性,导致智能体行为不可预测。
核心思路:GRUET通过将推理空间建模为图,精确量化轮次级推理的不确定性,并通过简单的聚合策略量化整体轨迹的可信度,从而提升智能体的可靠性。
技术框架:GRUET的整体架构包括两个主要模块:轮次级推理不确定性量化模块和轨迹级不确定性聚合模块。前者通过图复杂度近似推理空间复杂度,后者则通过聚合策略计算轨迹可信度。
关键创新:GRUET的创新在于将推理过程视为图结构,利用图的复杂度来量化推理不确定性,这一方法与传统的线性推理模型有本质区别。
关键设计:在设计中,采用了图的复杂度作为损失函数,确保推理空间的全面性和准确性,同时聚合策略简单有效,能够快速计算轨迹的整体可信度。
🖼️ 关键图片
📊 实验亮点
实验结果显示,GRUET在选择性生成性能上显著优于九个LLMs的基线,AUROC、AUPRC和AUARC指标均表现出明显提升,验证了该方法的有效性。
🎯 应用场景
该研究的潜在应用领域包括智能助手、自动驾驶、机器人决策等,能够提升智能体在复杂环境中的决策能力和可信度,具有重要的实际价值和未来影响。
📄 摘要(原文)
Agents have attracted considerably increasing attention due to the power of executing both Reasoning and Acting (ReAct) in open and dynamic environments. The ReAct process typically exhibits a multi-turn trajectory in which one drives Large Language Models (LLMs) to generate both reasoning chains and task-specific actions in an interleaved manner. However, agents often suffer from significant uncertainty, where identical tasks yield divergent trajectories; trajectories with higher uncertainty often produce incomprehensible behaviors, severely undermining agent credibility. This work conjectures that such trajectory-level uncertainty frequently stems from cumulative turn-level reasoning uncertainty induced by LLMs; the latter often exhibits a collection of branches of divergent reasoning chains and their resulting actions. Built upon this, we present the Graph-based Reasoning UncErtainty in Trajectories (GRUET) method for the uncertainty quantification of ReAct, comprising turn-level reasoning uncertainty quantification and trajectory-level uncertainty aggregation; the former precisely quantifies reasoning uncertainty via modeling the reasoning space spanned by potential reasoning branches as a graph and then approximating the reasoning space complexity with graph complexity, while the latter employs simple aggregation strategies for quantifying the overall trajectory credibility. Empirical evaluations across nine LLMs and five benchmarks validate the effectiveness of our proposed GRUET in terms of selective generation performance, measured by AUROC, AUPRC, and AUARC.