用视觉流缓解多智能体系统中的幻觉级联问题
Visual Multi-Agent System: Mitigating Hallucination Snowballing via Visual Flow
- 选取中间层视觉注意力峰值的特定视觉标记作为信息传递核心
- 实验显示在8个基准上显著降低幻觉级联，提升10种模型性能
- 轻量级插件方案，适配主流多智能体架构与视觉语言模型
基于视觉语言模型的多智能体系统虽能完成复杂任务，但面临新型失效模式——多智能体视觉幻觉级联：单个智能体产生的幻觉会因过度依赖文本流传递视觉信息而被后续智能体放大。通过逐轮、逐层、逐标记的注意力分析，我们揭示了幻觉级联的本质在于视觉注意力分配的减弱。研究发现，部分视觉标记在中间层具有单一模态注意力峰值，最能保留视觉证据，但在深层智能体轮次中逐渐衰减，导致幻觉传播。为此，我们提出ViF，一种轻量级、可即插即用的缓解方案，利用选定的视觉中继标记传递跨智能体消息，并通过注意力重分配强化该模式。实验结果表明，该方法显著减少幻觉级联，在基于四个常见多智能体结构和十种基础模型的八个基准上持续提升性能。源代码已公开于：https://github.com/YU-deep/ViF.git。
原文摘要 · Abstract (English)
Multi-Agent System (MAS) powered by Visual Language Models (VLMs) enables challenging tasks but suffers from a novel failure term, multi-agent visual hallucination snowballing, where hallucinations are seeded in a single agent and amplified by following ones due to the over-reliance on textual flow to relay visual information. Through turn-, layer-, and token-wise attention analyses, we provide detailed insights into the essence of hallucination snowballing regarding the reduction of visual attention allocation. It leads us to identify a subset of vision tokens with a unimodal attention peak in middle layers that best preserve visual evidence but gradually diminish in deeper agent turns, resulting in the visual hallucination snowballing in MAS. Thus, we propose ViF, a lightweight, plug-and-play mitigation paradigm that relays inter-agent messages with Visual Flow powered by the selected visual relay tokens and applies attention reallocation to amplify this pattern. The experiment results demonstrate that our method markedly reduces hallucination snowballing, consistently improving the performance across eight benchmarks based on four common MAS structures and ten base models. The source code is publicly available at: https://github.com/YU-deep/ViF.git.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。