GraphMAS: A Systematic Benchmark of Multi-Agent Coordination for Graph Learning

2026-09-30

AI Summary

This study addresses the lack of systematic exploration and evaluation of coordination mechanisms among LLM-based multi-agent systems for graph learning. We introduce the first shared expert pool, designing four paradigms encompassing seven methods along interaction and adaptation dimensions, alongside a unified evaluation protocol. Our experiments reveal the complementarity of heterogeneous perspectives and demonstrate that coordination capabilities transfer effectively to unseen graph tasks. Furthermore, we establish instance-level adaptive selection as the optimal strategy balancing performance against computational complexity. By systematically benchmarking multi-agent coordination in graph learning, this work provides critical insights and methodological foundations for advancing research in this direction.

Abstract

LLM-based multi-agent systems coordinate specialized reasoning through aggregation, interaction, and adaptive control, yet their potential for graph learning remains unexplored. Graph learning is a natural setting for such systems because useful evidence may arise from heterogeneous local, long-range, global structural, and semantic perspectives whose relevance varies across instances. Existing LLM-based graph learning approaches primarily rely on single-agent reasoning, while multi-agent coordination has been studied mainly in general reasoning settings. Consequently, it remains unclear whether multiple specialized agents can improve graph learning and how coordination strategies should be designed and evaluated. To address this gap, we introduce GraphMAS, a systematic benchmark of multi-agent coordination for graph learning. GraphMAS builds a shared pool of graph reasoning specialists and organizes coordination along two dimensions, inter-agent interaction and runtime adaptivity, yielding four paradigms and seven representative coordination methods. Under a unified protocol, we evaluate these methods across seven text-attributed graphs, three domains, and two graph learning tasks. We find that heterogeneous graph perspectives are complementary, and that coordinating specialists improves over individual specialists and single-agent graph reasoning, with gains from decomposing reasoning across specialists rather than from broader evidence access alone. However, richer inter-agent interaction does not reliably help, whereas instance-adaptive specialist selection yields the strongest accuracy-efficiency trade-off. We further show that coordination can be learned over a fixed specialist pool and transfers to held-out graphs. GraphMAS therefore provides a controlled evaluation framework and empirical principles for understanding when and how multi-agent coordination benefits graph learning.

Problem

• multi-agent coordination
• graph learning
• large language models
• benchmark

Innovation

• Multi-Agent Coordination
• Graph Learning
• Large Language Models
• Adaptive Specialist Selection
• Benchmark