LatticeMind：面向多智能体系统的冲突感知记忆原语

中文摘要 AI 辅助

多智能体大语言模型（LLM）系统的失败往往并非因为缺少候选答案，而是缺乏持久机制来决定当前应信任哪一个不相容的主张。多数投票、辩论和基于评判者的选择方法在选择输出时，不会记录哪个主张获胜、哪个存在争议，或后续更新为何会取代它。我们提出了LatticeMind，一种在写入时处理矛盾的冲突感知结构化记忆。它维护明确的条目状态，应用低成本的符号冲突检查，仅在未解决的语义案例中调用LLM进行协调。在移除了源名称提示的无标签ConflictBank评估中，LatticeMind达到了0.97的准确率，而最强的聚合基线仅为0.61，通过配对McNemar检验，该差距在p<10^-6时具有统计学显著性。 ablation研究显示，移除检查器或协调器会使准确率下降12至14个百分点。在四个次级规划基准上，结果喜忧参半：LatticeMind在四个基准中的三个上击败了朴素合并方法，但在奖励迭代搜索的任务中无法替代审议方法。

英文摘要

Multi-agent LLM systems often fail not for lack of candidate answers, but because they have no persistent mechanism for deciding which incompatible claim should currently be trusted. Majority vote, debate, and judge-based selection choose an output without recording which claim wins, which is contested, or why a later update supersedes it. We present \term{LatticeMind}, a conflict-aware structured memory that handles contradiction at write time. It maintains explicit item status, applies cheap symbolic conflict checks, and invokes LLM reconciliation only for unresolved semantic cases. On a label-blind ConflictBank evaluation that removes source-name hints, LatticeMind reaches 0.97 accuracy versus 0.61 for the strongest aggregation baseline, with the gap significant at $p<10^{-6}$ by paired McNemar test. Ablations show that removing the checker or the reconciler costs 12 to 14 points. On four secondary planning benchmarks the picture is mixed: LatticeMind beats naive merge on three of four, but does not replace deliberation methods on tasks rewarding iterative search.

发表机构

• University of Science and Technology of China（中国科学技术大学）
• University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
• Shanghai AI Laboratory（上海人工智能实验室）
• Beijing University of Posts and Telecommunications（北京邮电大学）
• University of Oxford（牛津大学）

机构由 AI 辅助整理，请以论文原文为准。