Token-Efficient Multi-Agent Collaboration via System One-Guided Computational Division of Labor
作者: Zihan Zhou, Xinzhe Hu, Hanxu Yang, Liangjian Wen, Zhao Kang
分类: cs.MA, cs.AI
发布日期: 2026-10-06
💡 一句话要点
提出S1-MAS以解决多代理系统中的协调效率问题
🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 多代理系统 大型语言模型 计算劳动分工 系统一模型 推理效率 协作性能 智能应用
📋 核心要点
- 现有的多代理系统在任务推理与协调操作之间存在紧密耦合,导致令牌开销和延迟显著增加,限制了可扩展性。
- S1-MAS框架通过将协调决策分配给轻量级的系统一模型,保留复杂推理给大型语言模型,从而实现高效的多代理协作。
- 在七个基准测试中,S1-MAS相比于现有方法在准确性上有显著提升,同时推理成本降低44.9%-97.2%,延迟减少37.8%-93.0%。
📝 摘要(中文)
基于大型语言模型的多代理系统(MAS)已成为复杂信息检索和推理任务的有前景的范式,然而现有框架将任务推理与协调操作紧密耦合,导致在交互增多时引入显著的令牌开销和延迟。本文提出S1-MAS,一个基于系统一指导的计算劳动分工的高效多代理框架,通过将协调决策分配给轻量级模型,保留开放式推理给强大的LLM工作者。实验表明,S1-MAS在七个基准测试中显著提高了准确性,同时减少了推理成本,展示了其在可扩展和经济高效的代理Web应用中的潜力。
🔬 方法详解
问题定义:本文旨在解决现有多代理系统在任务推理与协调操作之间的耦合问题,导致的令牌开销和延迟过高,限制了系统的可扩展性。
核心思路:S1-MAS通过将协调决策分配给轻量级的系统一模型,允许复杂推理留给大型语言模型,从而实现高效的计算劳动分工,提升协作性能。
技术框架:S1-MAS的整体架构包括一个轻量级控制器负责选择检查条件、后续任务和终止决策,以及一个紧凑的阅读器从授权源中检索相关证据,支持这些决策。
关键创新:S1-MAS的主要创新在于将协调与推理解耦,利用轻量级模型进行有限的协调决策,显著减少了推理过程中的令牌消耗和延迟。
关键设计:在设计中,控制器和阅读器的选择和配置是关键,确保在动态任务选择中能够有效地分配工作者角色和访问源,且无需针对特定任务进行训练。
🖼️ 关键图片
📊 实验亮点
实验结果显示,S1-MAS在七个基准测试中相比于AgentVerse、DyLAN和SelfOrg,减少了GPT-4o令牌消耗44.9%-97.2%,并且测得的端到端延迟降低了37.8%-93.0%,展现出显著的性能提升。
🎯 应用场景
该研究的潜在应用领域包括智能客服、自动化信息检索和复杂决策支持系统。通过提高多代理系统的效率,S1-MAS能够在资源有限的情况下提供更快速和准确的服务,具有广泛的实际价值和未来影响。
📄 摘要(原文)
Large language model (LLM)-based multi-agent systems (MAS) have become a promising paradigm for complex information-seeking and reasoning tasks by enabling collaborative problem solving among specialized agents. However, existing MAS frameworks tightly couple task reasoning with coordination operations, including task selection, role assignment, message routing, and context management. As interactions grow, using powerful LLMs for these bounded control decisions introduces substantial token overhead and latency, limiting the scalability of agentic Web services. In this paper, we investigate whether coordination can be decoupled from expensive reasoning without compromising collaborative performance. We propose S1-MAS, a token-efficient multi-agent framework based on System One-guided computational division of labor. S1-MAS assigns bounded coordination decisions to lightweight System One models while reserving open-ended reasoning for capable LLM workers. Specifically, a lightweight controller selects inspection conditions, chooses subsequent tasks, and determines termination, while a compact reader retrieves condition-relevant evidence from authorized sources to support these decisions. Through a decision-evidence loop, selected tasks dynamically determine worker roles and source access, enabling adaptive collaboration without task-specific training. Extensive experiments on seven diverse benchmarks demonstrate that S1-MAS achieves superior accuracy while substantially reducing the inference cost. Across individual comparisons with AgentVerse, DyLAN, and SelfOrg on seven benchmarks, S1-MAS reduces GPT-4o token consumption by 44.9%-97.2% and measured end-to-end latency by 37.8%-93.0%. These results highlight its potential for scalable and cost-effective agentic Web applications.