提出通过同步机制减少多智能体大模型因知识状态不同步导致的幻觉。
Hallucination as Context Drift: Synchronization Protocols for Multi-Agent LLM Systems
- 设计轻量级度量指标CDS和共享状态验证协议SSVP，定期校验智能体间知识状态差异。
- 实验显示SSVP使幻觉率降至0.463，比全广播同步降低34%，且节省58%调用次数。
- 发现幻觉污染效应仅出现在旅行规划等易受错误信念传播的任务中，具任务特异性。
多智能体大模型常产生无法仅由模型缺陷解释的幻觉输出。此类失败主要源于上下文漂移：并发智能体间内部知识状态的偏离。当智能体以不匹配或过时的共享世界状态进入协作任务时，联合推理会产生矛盾，表现为幻觉。本文提出上下文分歧得分（CDS），量化跨空间、时间与任务维度的智能体对间知识状态差异，并设计共享状态验证协议（SSVP），让智能体周期性交换压缩状态摘要，在联合推理前标记高分歧状态。在旅行规划与软件项目规划两个领域使用Claude Haiku进行评估。在控制实验中（旅行任务每条件n=30，软件任务n=10，共8场景），朴素全广播同步使幻觉率比无同步基线提高34%（HR: 0.658 vs. 0.492，p=0.0022，d=1.18），存在错误状态传播的污染效应。而SSVP避免该问题，实现小幅但稳定的幻觉率下降（HR: 0.463，d=0.30），显著优于全广播同步（p=0.0005，d=1.47），同时减少58%的API调用。污染效应在软件领域未复现，所有条件均收敛至低幻觉率（<0.2），表明其仅存在于一个错误信念可跨维度传播的任务中。结果将幻觉缓解重新定义为分布式系统问题，确立上下文同步为多智能体大模型设计的核心机制。
原文摘要 · Abstract (English)
Multi-agent LLM systems routinely produce hallucinated outputs that cannot be explained by model deficiencies alone. A significant class of these failures arises not from model incapacity but from context drift: the divergence of internal knowledge states between concurrent agents. When agents enter a collaborative task with mismatched or stale representations of shared world state, their joint reasoning produces contradictions that manifest as hallucination. We define the Context Divergence Score (CDS), a lightweight scalar metric quantifying knowledge-state discrepancy between agent pairs across spatial, temporal, and task dimensions, and propose the Shared State Verification Protocol (SSVP), which lets agents periodically exchange compressed state summaries and flag high-divergence conditions before joint reasoning. We evaluate SSVP across two domains (multi-agent travel and software project planning) using Claude Haiku. In controlled experiments (n=30 per condition, travel; n=10, software) across 8 scenarios, naive full-broadcast synchronization increases hallucination rate by 34% above the no-sync baseline (HR: 0.658 vs. 0.492, p=0.0022, d=1.18), a contamination effect from propagating erroneous agent states. SSVP avoids this failure mode while showing modest, consistent reduction (HR: 0.463, d=0.30) and achieves significantly lower hallucination than full-broadcast (p=0.0005, d=1.47) using 58% fewer API calls. The contamination effect does not replicate in the software domain, where all conditions converge to low HR (<0.2), confirming it is specific to tasks where one erroneous shared belief cascades across evaluation dimensions. Our results reframe hallucination mitigation as a distributed systems problem and establish context synchronization as a first-class primitive in multi-agent LLM design.
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。