Hear Both Sides: Efficient Multi-Agent Debate via Diversity-Aware Message Retention

TL;DR

This paper introduces Diversity-Aware Retention (DAR), a lightweight method for selecting diverse, disagreeing agent responses in multi-agent debates to improve reasoning quality and efficiency.

Contribution

DAR is a novel, index-based retention mechanism that preserves authentic disagreements, enhancing multi-agent debate performance especially with many agents.

Findings

DAR improves debate performance across various benchmarks.

Selective message propagation reduces noise and redundancy.

Performance gains increase with more agents.

Abstract

Multi-Agent Debate has emerged as a promising framework for improving the reasoning quality of large language models through iterative inter-agent communication. However, broadcasting all agent messages at every round introduces noise and redundancy that can degrade debate quality and waste computational resources. Current approaches rely on uncertainty estimation to filter low-confidence responses before broadcasting, but this approach is unreliable due to miscalibrated confidence scores and sensitivity to threshold selection. To address this, we propose Diversity-Aware Retention (DAR), a lightweight debate framework that, at each debate round, selects the subset of agent responses that maximally disagree with each other and with the majority vote before broadcasting. Through an explicit index-based retention mechanism, DAR preserves the original messages without modification, ensuring that retained disagreements remain authentic. Experiments on diverse reasoning and question answering benchmarks demonstrate that our selective message propagation consistently improves debate performance, particularly as the number of agents scales, where noise accumulation is most severe. Our results highlight that what agents hear is as important as what agents say in multi-agent reasoning systems. Code is publicly available at https://github.com/DA2I2-SLM/DAR.