Search papers, labs, and topics across Lattice.
This paper introduces R虏-MAD, a novel framework that enhances multi-agent debate by integrating experience memory and confidence estimation to address the shared misconception vulnerability. By employing a debate-state-aware retrieval policy, R虏-MAD dynamically adjusts agents' concept priors based on historical evidence, while also estimating per-agent reliability to modulate influence during debates. Experimental results demonstrate that R虏-MAD significantly outperforms both single-agent and existing multi-agent debate baselines across various benchmarks, highlighting its effectiveness in refining reasoning capabilities in large language models.
R虏-MAD not only corrects misconceptions in multi-agent debates but also intelligently weighs agent contributions based on past performance, leading to more accurate outcomes.
Multi-agent debate (MAD) improves the reasoning capabilities of large language models by having multiple agents iteratively refine their responses through discussion. However, MAD suffers from a critical vulnerability known as shared misconception: when a majority of agents initially converge on an incorrect answer, the debate process tends to amplify rather than correct the error. Existing methods primarily address peer skew but leave the agents'inherently biased concept priors unaddressed. To mitigate this systematic weakness, we propose R$^2$-MAD (Remember and Reweight for Multi-Agent Debate), a framework that equips agents with an experience memory accumulated from past debates. R$^2$-MAD intervenes on both failure modes through two complementary mechanisms: A debate-state-aware retrieval policy dynamically calibrates the concept prior by retrieving relevant historical evidence based on the current consensus level. Then these retrieved experiences provide a basis for estimating per-agent reliability, yielding confidence weights to modulate peer influence. Experiments on various benchmarks show that R$^2$-MAD achieves consistent improvements over existing single-agent and MAD baselines.
