RADAR: Redundancy-Aware Diffusion for Multi-Agent Communication Structure Generation

TL;DR

RADAR is a novel framework that dynamically generates communication structures for multi-agent systems, reducing redundancy and improving performance across various tasks.

Contribution

It introduces a step-by-step, redundancy-aware diffusion model for adaptive communication topology generation in multi-agent systems.

Findings

RADAR outperforms recent baselines in accuracy and robustness.

It achieves lower token consumption while maintaining high performance.

Experimental results are validated across six diverse benchmarks.

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.