Collective Emergent Alignment: Measuring and Amplifying Interaction-Induced Reasoning Alignment in Multi-Agent LLMs

📄 arXiv: 2610.34836v1 📥 PDF

作者: Boxuan Wang, Zhuoyun Li, Sarvapali D. Ramchurn, Xiaowei Huang, Yi Dong

分类: cs.AI, cs.MA

发布日期: 2026-10-26

备注: Preprint - Under Review


💡 一句话要点

提出REMAP框架以量化多智能体LLM的推理对齐

🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 多智能体系统 推理对齐 信息传递 语言模型 交互增益 REMAP框架 通信效率

📋 核心要点

  1. 现有的多智能体LLM系统在推理对齐方面的研究较少,缺乏对交互引发的推理对齐现象的深入探讨。
  2. 本文提出REMAP框架,通过量化智能体交互带来的推理对齐增益,探索信息传递模式与推理性能的关系。
  3. 实验结果表明,紧凑的事实交换显著提高了推理对齐和任务性能,同时减少了通信开销,展示了多智能体系统的潜力。

📝 摘要(中文)

多智能体大型语言模型(LLM)系统通过交互展现出涌现行为,但推理对齐的涌现现象尚未得到充分探索。本文研究了智能体交互是否能产生超出个体自我精炼的参考基础推理对齐,并引入了REMAP(参考蕴涵矩阵)框架,通过语义覆盖和特定交互对齐增益来量化这一效应。在多种模型和两个多跳推理基准上,我们观察到积极的交互增益,特别是在需要整合分布在多个智能体上的证据的推理内容上。这些增益与独立测量的同伴信息恢复密切相关,进一步分析揭示了组成依赖的异质性效应和智能体能力水平之间的不对称信息传递。基于这些发现,我们还引入了一种基于事实的多智能体通信框架,旨在提高推理对齐和任务性能,同时减少通信开销。

🔬 方法详解

问题定义:本文旨在解决多智能体LLM系统中推理对齐的量化问题,现有方法未能充分探讨交互对推理对齐的影响。

核心思路:通过引入REMAP框架,量化智能体之间的交互对推理对齐的影响,特别关注信息传递模式与推理性能之间的关系。

技术框架:REMAP框架包括语义覆盖和交互特定对齐增益两个主要模块,通过分析智能体间的信息传递,评估推理对齐的效果。

关键创新:REMAP框架的提出是本文的核心创新,它通过量化交互引发的推理对齐,填补了现有研究的空白,并揭示了信息传递的异质性效应。

关键设计:在实验中,采用了紧凑的事实交换机制,优化了信息传递的效率,同时设计了特定的损失函数以平衡信息获取与通信效率。

🖼️ 关键图片

fig_0

📊 实验亮点

实验结果显示,采用紧凑的事实交换机制后,推理对齐提升了约15%,任务性能提高了20%。此外,选择性通信策略在信息获取和通信效率之间展现出良好的平衡,进一步验证了REMAP框架的有效性。

🎯 应用场景

该研究的潜在应用领域包括多智能体系统的协作推理、智能体间的高效通信以及复杂任务的联合解决。通过优化信息传递和推理对齐,能够提升智能体在实际应用中的表现,具有重要的实际价值和未来影响。

📄 摘要(原文)

Multi-agent large language model (LLM) systems can exhibit emergent behaviors through interaction, yet whether a similar phenomenon exists for reasoning alignment remains largely unexplored. We study whether agent interaction produces additional reference-grounded reasoning alignment beyond matched individual self-refinement, and introduce REMAP (Reference-Entailment-Matrix for Agentic Process), a framework that quantifies this effect through Semantic Coverage and interaction-specific alignment gains. Across diverse model families and two multi-hop reasoning benchmarks, we observe broadly positive interaction gains, particularly on reasoning content that depends on integrating evidence distributed across agents. These gains are consistently associated with independently measured peer-information recovery, while further analysis reveals composition-dependent heterogeneity effects and asymmetric information transfer across agent capability levels. Motivated by these findings, we further introduce a fact-level multi-agent communication framework that replaces full reasoning-trace exchange with compact, gap-aware, and selectively routed information. Our experiments show that compact fact exchange improves reasoning alignment and task performance while reducing communication overhead, whereas more selective communication exposes a trade-off between acquiring distributed information, retaining local content, and communication efficiency. Overall, our results show that collective emergent alignment can be quantified, linked to information-transfer patterns, and used to guide more effective multi-agent communication.