Finding Where the Buck Stops: An Automated Failure Attribution-Based Reflection Framework for Multi-Agent Collaboration

📄 arXiv: 2608.28264v1 📥 PDF

作者: Xiaoqing Wang, Keman Huang, Bin Liang, Hongyu Li, Xiaoyong Du, Wuqiong Pan

分类: cs.AI

发布日期: 2026-08-28

备注: Accepted by EMNLP 2026 main


💡 一句话要点

提出DoCtOR框架以解决多智能体系统中的失败归因问题

🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture) 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 多智能体系统 失败归因 反思机制 自动化推理 PPO优化 协作机器人 智能决策

📋 核心要点

  1. 现有的多智能体系统反思方法未能有效识别导致失败的关键智能体,导致反思过程中的信息污染。
  2. 本文提出的DoCtOR框架通过自动化失败归因,专注于决策错误智能体的反思,避免了其他智能体的错误信息干扰。
  3. 实验结果显示,DoCtOR在多个数据集上显著提升了成功率,验证了其有效性和普适性。

📝 摘要(中文)

多智能体系统(MAS)在复杂任务中展现出潜力,但面临高失败率的挑战。现有的自我反思方法要求所有智能体对失败进行反思,忽视了失败通常源自特定智能体的事实。为此,本文提出DoCtOR(Diagnose-then-Correct PPO-enhanced Reflection),通过自动化失败归因识别关键错误步骤和决策错误智能体,利用反事实推理生成修正步骤,并仅让决策错误智能体进行针对性反思。实验结果表明,DoCtOR在HotPotQA、ChartQAPro和Mind2Web数据集上分别提升了22%、26%和27%的成功率,超越了现有方法如Reflexion、Retroformer和COPPER。

🔬 方法详解

问题定义:本文旨在解决多智能体系统中失败归因不准确的问题。现有方法要求所有智能体反思,导致信息污染和反思效率低下。

核心思路:DoCtOR框架通过自动化识别决策错误智能体,利用反事实推理生成修正步骤,专注于关键智能体的反思,从而提高反思的有效性。

技术框架:DoCtOR的整体流程包括三个主要阶段:首先,自动化失败归因识别关键错误步骤和决策错误智能体;其次,利用反事实推理生成修正的决策错误步骤;最后,仅让决策错误智能体进行针对性反思。

关键创新:DoCtOR的创新在于其“诊断-修正”范式,区别于传统方法的全面反思,聚焦于特定智能体的反思,避免了信息污染。

关键设计:在技术细节上,DoCtOR采用了PPO(Proximal Policy Optimization)增强的反思机制,优化了反思过程中的损失函数设置,以提高反思的针对性和有效性。

🖼️ 关键图片

fig_0
fig_1

📊 实验亮点

实验结果表明,DoCtOR在HotPotQA、ChartQAPro和Mind2Web数据集上分别实现了22%、26%和27%的成功率提升,显著优于基线方法如Reflexion、Retroformer和COPPER,验证了其有效性和优势。

🎯 应用场景

该研究的潜在应用领域包括智能客服、自动驾驶和协作机器人等多智能体系统。通过提高反思效率和准确性,DoCtOR能够显著提升系统的整体性能和可靠性,具有重要的实际价值和未来影响。

📄 摘要(原文)

Multi-agent systems (MAS) powered by large language models have shown promise for complex tasks but suffer from high failure rates. Current self-reflection methods for MAS require all agents to reflect upon failure, overlooking a critical reality: failures typically stem from a specific agent leading the task astray, namely the decisive error agent, while others merely fulfill their regular duties. Forcing regular-behaving agents to reflect contaminates their memory with wrong insights. Hence, we propose DoCtOR (Diagnose-then-Correct PPO-enhanced Reflection), a novel reflection framework that enhances multi-agent collaboration. DoCtOR first identifies the decisive error step and decisive error agent through automated failure attribution, then employs counterfactual reasoning to generate a corrected decisive error step, and finally engages only the decisive error agent to produce targeted reflections. Experimental results show DoCtOR achieves 22%, 26%, and 27% improvements over initial success rates on HotPotQA, ChartQAPro, and Mind2Web datasets, outperforming Reflexion, Retroformer, and COPPER. We further establish the generalizability of our diagnose-then-correct paradigm and demonstrate that in low-resource settings, focusing reflection on reasoning steps after the decisive error step achieves comparable quality to reflecting on the complete failure trajectory.