OFA-MAS: One-for-All Multi-Agent System Topology Design based on Mixture-of-Experts Graph Generative Models

TL;DR

OFA-MAS introduces a universal framework for designing adaptive multi-agent system topologies from natural language descriptions, outperforming specialized models by leveraging a mixture-of-experts architecture and multi-stage training.

Contribution

It presents the first one-for-all model that generates MAS topologies from natural language, integrating a task-aware encoder and mixture-of-experts for improved adaptability.

Findings

Outperforms specialized models on six benchmarks.

Generates highly adaptive and task-relevant topologies.

Effective across diverse cross-domain scenarios.

Abstract

Multi-Agent Systems (MAS) offer a powerful paradigm for solving complex problems, yet their performance is critically dependent on the design of their underlying collaboration topology. As MAS become increasingly deployed in web services (e.g., search engines), designing adaptive topologies for diverse cross-domain user queries becomes essential. Current graph learning-based design methodologies often adhere to a "one-for-one" paradigm, where a specialized model is trained for each specific task domain. This approach suffers from poor generalization to unseen domains and fails to leverage shared structural knowledge across different tasks. To address this, we propose OFA-TAD, a one-for-all framework that generates adaptive collaboration graphs for any task described in natural language through a single universal model. Our approach integrates a Task-Aware Graph State Encoder (TAGSE) that filters task-relevant node information via sparse gating, and a Mixture-of-Experts (MoE) architecture that dynamically selects specialized sub-networks to drive node and edge prediction. We employ a three-stage training strategy: unconditional pre-training on canonical topologies for structural priors, large-scale conditional pre-training on LLM-generated datasets for task-topology mappings, and supervised fine-tuning on empirically validated graphs. Experiments across six diverse benchmarks show that OFA-TAD significantly outperforms specialized one-for-one models, generating highly adaptive MAS topologies. Code: https://github.com/Shiy-Li/OFA-MAS.