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

2026-05-10

AI Summary

Existing multi-agent systems often rely on fixed or single-step generated communication topologies, which struggle to balance resource efficiency for simple tasks with collaborative capacity for complex ones. This work proposes a redundancy-aware, query-adaptive dynamic topology generation framework that, for the first time, introduces a conditional discrete graph diffusion model into multi-agent communication structure optimization. By leveraging a graph effective size–guided progressive generation mechanism, the method enables fine-grained, efficient, and adaptive topology construction. Evaluated across six benchmark tasks, the approach significantly outperforms existing methods, achieving higher accuracy while reducing communication overhead and demonstrating enhanced robustness across diverse scenarios.

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.

Problem

• multi-agent communication
• communication topology
• token efficiency
• structural exploration
• redundancy

Innovation

• redundancy-aware diffusion
• multi-agent communication
• graph generation
• conditional discrete diffusion
• query-adaptive topology