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

2026-03-20

AI Summary

This work addresses the inefficiency and high communication overhead in existing large language model (LLM)-based multi-agent systems, which stem from the lack of explicit modeling of task-oriented collaborative groups in communication topology construction. To overcome this limitation, the authors propose GoAgent, a novel approach that treats collaborative groups as the fundamental atomic units of the communication topology. GoAgent leverages an LLM to enumerate candidate groups and autoregressively connects them to form a communication graph. Furthermore, it incorporates a Conditional Information Bottleneck (CIB) mechanism to compress inter-group communication and suppress noise propagation. Evaluated across six benchmark tasks, GoAgent achieves an average accuracy of 93.84% while reducing token consumption by approximately 17%, demonstrating significant improvements in both communication efficiency and collaborative performance.

Technology Category

• Multiagent Systems: Agent Communication
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

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%.

Problem

• multi-agent systems
• communication topology
• group structure
• LLM-based coordination
• task decomposition

Innovation

• Group-of-Agents
• Communication Topology Generation
• Conditional Information Bottleneck
• LLM-based Multi-Agent Systems
• Autoregressive Group Selection