GoAgent: Group-of-Agents Communication Topology Generation for LLM-based Multi-Agent Systems
arXiv:2603.19677v1 Announce Type: new 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 an
arXiv:2603.19677v1 Announce Type: new 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%.
Executive Summary
The article proposes GoAgent, a communication topology generation method for Large Language Model (LLM)-based Multi-Agent Systems (MAS). GoAgent explicitly models collaborative groups as atomic units, addressing the limitation of existing approaches that generate topologies in a node-centric manner. The method first enumerates task-relevant candidate groups using an LLM and then autoregressively selects and connects these groups. To mitigate communication redundancy and noise propagation, GoAgent introduces a conditional information bottleneck (CIB) objective. Experiments on six benchmarks demonstrate the state-of-the-art performance of GoAgent with 93.84% average accuracy while reducing token consumption by 17%. GoAgent's innovative approach has significant implications for improving coordination and reducing unnecessary communication overhead in MAS.
Key Points
- ▸ GoAgent proposes a novel communication topology generation method for LLM-based MAS
- ▸ The method explicitly models collaborative groups as atomic units
- ▸ GoAgent introduces a conditional information bottleneck (CIB) objective to mitigate communication redundancy and noise propagation
Merits
Strength
GoAgent's explicit modeling of collaborative groups addresses a significant limitation of existing approaches, leading to improved coordination and reduced unnecessary communication overhead.
Demerits
Limitation
The method relies heavily on the quality and accuracy of the LLM used to enumerate task-relevant candidate groups, which may introduce bias or variability in the generated topologies.
Expert Commentary
The article makes a significant contribution to the field of Multi-Agent Systems by proposing a novel communication topology generation method that explicitly models collaborative groups. The use of a conditional information bottleneck (CIB) objective is a clever approach to mitigating communication redundancy and noise propagation. However, the method's reliance on the quality and accuracy of the LLM used to enumerate task-relevant candidate groups is a limitation that requires careful consideration. Overall, GoAgent's innovative approach has significant implications for improving coordination and reducing unnecessary communication overhead in MAS, and its development has important practical and policy implications.
Recommendations
- ✓ Future research should focus on developing more sophisticated and accurate methods for modeling complex tasks and collaborative behaviors in MAS, particularly in areas such as robotics, autonomous systems, and smart cities.
- ✓ Developers of MAS should carefully consider the responsible development and deployment of these systems, particularly with regards to issues of transparency, explainability, and accountability.
Sources
Original: arXiv - cs.LG
