DyTopo: Dynamic Topology Routing for Multi-Agent Reasoning via Semantic Matching

TL;DR

DyTopo is a dynamic multi-agent communication framework that improves reasoning by adaptively routing messages based on semantic matching, leading to better performance and interpretability across various benchmarks.

Contribution

Introduces DyTopo, a novel manager-guided framework that reconstructs sparse communication graphs at each round using semantic matching, enhancing multi-agent reasoning.

Findings

Outperforms baseline by an average of +6.2 in accuracy.

Provides interpretable communication graphs for qualitative analysis.

Consistently improves performance across multiple benchmarks and LLM backbones.

Abstract

Multi-agent systems built from prompted large language models can improve multi-round reasoning, yet most existing pipelines rely on fixed, trajectory-wide communication patterns that are poorly matched to the stage-dependent needs of iterative problem solving. We introduce DyTopo, a manager-guided multi-agent framework that reconstructs a sparse directed communication graph at each round. Conditioned on the manager's round goal, each agent outputs lightweight natural-language query (need) and \key (offer) descriptors; DyTopo embeds these descriptors and performs semantic matching, routing private messages only along the induced edges. Across code generation and mathematical reasoning benchmarks and four LLM backbones, DyTopo consistently outperforms over the strongest baseline (avg. +6.2). Beyond accuracy, DyTopo yields an interpretable coordination trace via the evolving graphs, enabling qualitative inspection of how communication pathways reconfigure across rounds.