arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2603.19677cs.LGcs.AIcs.MA
收藏

GoAgent:基于大语言模型的多智能体系统的群体通信拓扑生成

GoAgent: Group-of-Agents Communication Topology Generation for LLM-based Multi-Agent Systems

School of Cyberspace, Hangzhou Dianzi University(杭州电子科技大学信息学院);School of Computing Technologies, RMIT University(皇家墨尔本理工大学计算技术学院);School of Information and Communication Technology, Griffith University(格里菲斯大学信息与通信技术学院)

Hongjiang Chen, Xin Zheng, Yixin Liu, Pengfei Jiao, Shiyuan Li, Huan Liu, Zhidong Zhao, Ziqi Xu, Ibrahim Khalil, Shirui Pan

AI总结:

本文提出GoAgent,通过LLM生成任务相关的群体结构,显式构建多智能体系统通信拓扑,提升协作效率并减少通信开销。

摘要:

基于大语言模型(LLM)的多智能体系统(MAS)在解决复杂任务方面表现出色,但其效果依赖于协调智能体交互的底层通信拓扑。在这些系统中,成功的问题解决往往需要任务特定的群体结构来分解和解决子任务。然而,大多数现有方法以节点为中心生成通信拓扑,使群体结构隐含地从局部连接决策中出现,而非显式建模,导致协调不充分和不必要的通信开销。为了解决这一限制,我们提出了GoAgent(群体智能体),一种通信拓扑生成方法,将协作群体作为MAS构建的原子单位。具体而言,GoAgent首先通过LLM枚举任务相关的候选群体,然后自回归地选择和连接这些群体作为原子单位来构建最终的通信图,共同捕捉组内凝聚力和组间协调。为了减轻拓扑扩展中固有的通信冗余和噪声传播,我们进一步引入了一个条件信息瓶颈(CIB)目标,压缩组间通信,保留任务相关的信号,同时过滤掉冗余的历史噪声。在六个基准上的广泛实验表明,GoAgent在平均准确率为93.84%的同时,将token消耗减少了约17%。

英文摘要:

Large language model (LLM)-based multi-agent systems (MAS) have demonstrated exceptional capabilities in solving complex tasks, yet their effectiveness depends heavily on the underlying communication topology that coordinates agent interactions. Within these systems, successful problem-solving often necessitates task-specific group structures to divide and conquer subtasks. However, most existing approaches generate communication topologies in a node-centric manner, leaving group structures to emerge implicitly from local connectivity decisions rather than modeling them explicitly, often leading to suboptimal coordination and unnecessary communication overhead. To address this limitation, we propose GoAgent (Group-of-Agents), a communication topology generation method that explicitly treats collaborative groups as the atomic units of MAS construction. Specifically, GoAgent first enumerates task-relevant candidate groups through an LLM and then autoregressively selects and connects these groups as atomic units to construct the final communication graph, jointly capturing intra-group cohesion and inter-group coordination. To mitigate communication redundancy and noise propagation inherent in expanding topologies, we further introduce a conditional information bottleneck (CIB) objective that compresses inter-group communication, preserving task-relevant signals while filtering out redundant historical noise. Extensive experiments on six benchmarks demonstrate the state-of-the-art performance of GoAgent with 93.84% average accuracy while reducing token consumption by about 17%.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑