Multi agent graph learning improves reasoning on complex graphs
Multi-Agent Agentic Graph Learning via Structural Signatures
Artificial Intelligence
Summary
Graphs can be tricky because different parts may have very different patterns. The authors designed a way for multiple AI agents to work together by splitting the graph into communities. Each agent focuses on its own part, using special compact summaries of graph structure that don’t depend on how the graph is described. These agents also check when they need help and debate to improve their answers. This teamwork helped them do better on standard tests than earlier single-agent methods.
What this means in practice
- •For knowledge graph engineers: Improve automated reasoning over complex knowledge graphs by distributing tasks among specialized agents that collaborate through compact structural summaries.
- •For cybersecurity analysts: Detect and analyze complex network patterns by assigning agents to subnetworks that summarize structure efficiently and debate on suspicious activities.
Authors
Liang Qu, Jianxin Li, Hua Wang
Abstract
Agentic graph learning (AGL) has recently achieved promising results on graph reasoning tasks, where an agent powered by a large language model (LLM) sequentially samples the graph as evidence to support its final prediction. Existing methods either employ a single agent or orchestrate multiple role-based agents to reason and learn over the entire graph, but both essentially rely on a shared reasoning policy across different graph regions, which can be suboptimal for graphs with heterogeneous structural and semantic patterns. Inspired by the progress of multi-agent collaboration on complex reasoning tasks, a natural remedy is to let multiple agents own different memory and collaborate; however, applying this paradigm to graphs directly faces two challenges. First, existing AGL methods typically verbalize graph structures into natural-language descriptions for LLM agents, making the reasoning process sensitive to the ordering of structural information and thereby breaking the permutation-invariant nature of graphs. Second, incorporating increasingly large sampled neighborhoods leads to rapidly growing contexts. To address these challenges, this paper introduces a multi-agent agentic graph learning (i.e., MAAGL) framework. MAAGL partitions the graph into communities and assigns an independent agent to each community for region-specific specialization. MAAGL represents structural and semantic evidence separately. Structural evidence is summarized by a dynamically updated structural signature that is permutation-invariant and fixed in size, while semantic evidence is filtered to the top-k nodes ranked by relevance. Based on historical trajectories with similar signatures, agents estimate their confidence and trigger debate-style collaboration when needed. Extensive experiments on four benchmark datasets show that MAAGL outperforms SOTA AGL methods.