Epistemic Gain, Aleatoric Cost: Uncertainty Decomposition in Multi-Agent Debate for Math Reasoning

2026-03-01

AI Summary

This work addresses the unclear mechanisms by which information exchange in multi-agent debate (MAD) influences reasoning capabilities, particularly in light of observed paradoxes such as accuracy gains accompanied by entropy increase and divergent performance between homogeneous and heterogeneous model ensembles. Introducing, for the first time, a Bayesian uncertainty decomposition perspective, the study disentangles total predictive uncertainty into epistemic uncertainty—reducible through debate—and aleatoric uncertainty stemming from intrinsic model noise. Building on this decomposition, the authors propose an uncertainty-guided multi-agent reinforcement learning algorithm that optimizes the trade-off between high epistemic gain and controllable aleatoric cost. This approach significantly enhances post-debate accuracy, stability, and individual reasoning performance, outperforming existing single-agent reinforcement learning methods.

Technology Category

• Reasoning under Uncertainty: Uncertainty Representations
• Multiagent Systems: Multiagent Systems under Uncertainty
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Multi-Agent Debate (MAD) has shown promise in leveraging collective intelligence to improve reasoning and reduce hallucinations, yet it remains unclear how information exchange shapes the underlying ability. Empirically, MAD exhibits paradoxical phenomena, such as accuracy improvement accompanied by substantial increase in token entropy, and remarkable divergence between homogeneous and heterogeneous model combinations. In this paper, we propose a Bayesian uncertainty analysis framework for MAD, which decomposes total predictive uncertainty into epistemic uncertainty reducible by debate context and aleatoric uncertainty induced by internal model noise. Across multiple model configurations, we find that effective debate hinges on achieving high epistemic gain under controlled aleatoric cost. Building on this insight, we design an uncertainty-guided multi-agent reinforcement learning (MARL) algorithm that explicitly optimizes aleatoric noise reduction and epistemic information utilization. Experiments show that our training significantly improves post-debate accuracy and stability, and enhances individual reasoning beyond single-agent RL, providing a unified Bayesian uncertainty perspective for understanding and improving MAD.

Problem

• Multi-Agent Debate
• Uncertainty Decomposition
• Epistemic Uncertainty
• Aleatoric Uncertainty
• Math Reasoning

Innovation

• Uncertainty Decomposition
• Epistemic Uncertainty
• Aleatoric Uncertainty
• Multi-Agent Debate
• Bayesian Reinforcement Learning