RADAR: Redundancy-Aware Diffusion for Multi-Agent Communication Structure Generation
2026-05-11 • Artificial Intelligence
Artificial IntelligenceMultiagent Systems
AI summaryⓘ
The authors studied how groups of large language models work together to solve tasks and found that their fixed ways of communicating can be inefficient or limiting. They created RADAR, a method that changes communication step-by-step based on what is needed, cutting down unnecessary messaging. Tests showed that RADAR makes these model groups more accurate, uses fewer words, and works better across different types of problems. This approach helps multi-agent systems communicate more thoughtfully and effectively.
large language modelsmulti-agent systemscommunication topologygraph diffusion modelstoken efficiencygenerative frameworkconditional generationrobustnesscode generationmathematical reasoning
Authors
Zhen Zhang, Wanjing Zhou, Juncheng Li, Hao Fei, Jun Wen, Wei Ji
Abstract
Compared with individual agents, large language model based multi-agent systems have shown great capabilities consistently across diverse tasks, including code generation, mathematical reasoning, and planning, etc. Despite their impressive performance, the effectiveness and robustness of these systems heavily rely on their communication topology, which is often fixed or generated in a single step. This restricts fine-grained structural exploration and flexible composition, resulting in excessive token utilization on simple tasks while limiting capability on complicated tasks. To mitigate this challenge, we introduce RADAR, a redundancy-aware and query-adaptive generative framework that actively reduce communication overhead. Motivated by recent progress in conditional discrete graph diffusion models, we formulate communication topology design as a step-by-step generation process, guided by the effective size of the graph. Comprehensive experiments on six benchmarks demonstrate that RADAR consistently outperforms recent baselines, achieving higher accuracy, lower token consumption, and greater robustness across diverse scenarios. Our code and data are available at https://github.com/cszhangzhen/RADAR.