多智能体对话中共识易成幻觉，关键事实与观点会逐渐消失。
The Deliberative Illusion: Diagnosing Factual Attrition and Stance Homogenization in Multi-Agent LLM Deliberation
- 通过分解问题为原子事实，追踪其在讨论中的存亡
- 高达72%的关键事实在讨论中丢失，导致信息严重缩水
- 适合关注大模型协作可靠性与偏见传播的研究者
多智能体大模型系统常将共识视为有效交互的证据。然而，在审议类问题中，可靠性取决于智能体是否保留解读议题所需的事实与立场。我们识别出‘审议幻觉’：讨论过程伴随（1）事实衰减——关键事实逐步流失，以及（2）立场同质化——多元观点趋同于统一结论。为此提出DelibTrace框架，将每个议题分解为原子事实，标注关键事实，分发至各智能体，并追踪其在多轮讨论中的存活率。在三类主流大模型家族对伦理与新闻类议题的实验中，关键事实最多损失达72%。该损失具有实质性影响：残留证据可误导性重构议题，最终立场仍受基础模型先验锚定，且单一恶意智能体可向不断缩小的共享语境注入虚假信息。结果揭示更深层风险：智能体可能越一致，却越无知。我们呼吁评估应关注哪些事实、不确定性与合理分歧得以留存。
原文摘要 · Abstract (English)
Multi-agent LLM systems often treat consensus as evidence of successful interaction. For deliberative problems, however, reliability depends on whether agents preserve the facts and viewpoints needed to interpret an issue. We identify the deliberative illusion: discussion produces (1) factual attrition, the progressive loss of issue-critical facts, alongside (2) stance homogenization, the collapse of diverse positions toward consensus. To measure this process, we introduce DelibTrace, a framework that decomposes each issue into atomic facts, labels issue-critical ones, distributes them across agents, and tracks their survival across discussion rounds. Across ethical and news-based deliberation with three representative LLM families, multi-agent discussion erases up to 72% of issue-critical facts. This loss is consequential: retained evidence can reconstruct the issue misleadingly, final stances remain anchored in base-model priors, and a single malicious agent can inject misinformation into the shrinking shared context. These results reveal a sharper risk: agents can agree more while knowing less. We call for evaluations that measure which facts, uncertainties, and legitimate disagreements survive interaction.
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。