FlowMAS: Learning Multi-Agent Workflow Topology via Information-guided Generative Flow Network

2026-09-29

AI Summary

This study addresses the high computational cost, coarse-grained feedback, and difficulty in modeling complex discrete dependencies inherent in existing multi-agent workflow topology generation methods. To overcome these limitations, this work proposes a topology optimization framework based on Generative Flow Networks (GFlowNets). Specifically, it introduces a novel structure-aware exploration mechanism coupled with an information contribution evaluation strategy. By integrating curiosity-driven policies with large language models (LLMs), the proposed approach enables efficient discrete topology search and collaboration pattern optimization. Extensive experiments across six benchmark datasets and three LLM backbones demonstrate that our method consistently outperforms existing baselines. Ultimately, this work establishes a new paradigm for the automated design of multi-agent systems.

Abstract

Automated multi-agent systems offer clear advantages over manually designed ones in scalability and adaptability, but existing workflow topology methods still face important limitations. Search-based methods are often computationally expensive, textual-gradient-based methods rely on coarse-grained feedback, and existing generation-based methods are not well suited to discrete workflow topologies with complex dependencies. To address these limitations, we propose FlowMAS, a multi-agent workflow topology method based on Generative Flow Networks (GFlowNets). FlowMAS models workflow generation as reward-guided flow over the topology space and introduces three components: a GFlowNet-based topology generation backbone, a curiosity-driven module for structure-aware exploration, and an information-guided optimization module for evaluating intermediate topologies. Concretely, the curiosity-driven module encourages exploration of structurally novel workflows, while the information-guided module measures both the information contribution and the communication efficiency of different operators to favor more informative and effective collaboration patterns. Experiments on six benchmark datasets with three LLM backbones show that FlowMAS consistently outperforms multiple baselines.

Problem

• Multi-Agent Systems
• Workflow Topology
• Generative Flow Networks
• Discrete Optimization
• Large Language Models

Innovation

• Generative Flow Networks
• Multi-Agent Workflow Topology
• Curiosity-driven Exploration
• Information-guided Optimization
• Automated Multi-Agent Systems