Discovering Efficient and Explainable Communication Topologies for LLM-based Multi-Agent Systems via Causal Inference

2026-08-13

AI Summary

Current multi-agent systems based on large language models lack an interpretable mechanism for generating communication topologies, making it difficult to identify subgraphs critical to collaborative success. This work proposes the E2-Explainer framework, which introduces causal reasoning into communication topology explanation for the first time by formulating it as an edge-level causal attribution problem. The approach identifies compact, influential subgraphs through task-preserving evidence and response stability metrics, and employs an amortized explainer to enable efficient post-hoc interpretation. Being model-agnostic, the method can be directly applied to communication pruning, significantly reducing communication overhead while maintaining competitive task performance across multiple reasoning and programming benchmarks.

Abstract

The performance of large language model (LLM)-based multi-agent systems (MAS) largely depends on effective communication topologies. Existing topology generation methods, however, typically learn communication topologies through black-box optimization driven solely by task-level rewards. While effective, such optimization provides little insight into why particular communication edges are selected, making it difficult to identify the critical communication subgraphs responsible for successful collaboration. To address this limitation, we propose E2-Explainer, a model-agnostic framework for providing interpretable explanations of communication topologies produced by arbitrary topology generators. Specifically, we formulate topology explanation as a causal attribution problem that identifies compact communication subgraphs supported by edge-level evidence of task preservation. We obtain this evidence with a Granger-style objective that measures how masking each communication channel changes the task outcome and the stability of the final response. The resulting budgeted subgraphs are then distilled into an amortized explainer, enabling efficient post-hoc explanation without repeated edge-level evaluations at deployment. Extensive experiments on multiple reasoning and coding benchmarks demonstrate that E2-Explainer identifies critical communication subgraphs that preserve successful collaboration. These subgraphs can also be executed directly to prune redundant communication edges, substantially reducing communication costs while maintaining competitive task performance.

Problem

• communication topologies
• explainability
• multi-agent systems
• causal inference
• LLM-based systems

Innovation

• causal inference
• explainable AI
• communication topology
• multi-agent systems
• LLM-based reasoning