Sequential Probabilistic Uncertainty Estimation for Parallel Multi-Agent Reasoning Systems

2026-10-06

AI Summary

This study addresses the challenge of inadequate uncertainty estimation in parallel multi-agent reasoning systems by proposing SAUCE, a novel method that introduces the first consensus-evolution-based sequential Bayesian filtering framework. By formulating uncertainty as a sequential inference process, SAUCE employs particle-filter-style updates to aggregate cross-round consistency and generation signals, enabling lightweight, training-free collaborative uncertainty estimation among multiple agents. Experimental evaluations across five benchmarks and two protocols demonstrate that SAUCE significantly outperforms conventional log-likelihood approaches and specialized baselines, effectively enhancing both misclassification detection and calibration performance.

Abstract

LLM-based multi-agent systems (MAS) have attracted growing attention for improving reasoning through interaction among multiple agents. In this work, we focus on parallel multi-agent reasoning systems, where several agents solve the same problem over multiple rounds and aggregate their outputs into a final answer. Despite their strong reasoning performance, uncertainty estimation for such systems remains underexplored: the reliability of a MAS depends not only on individual generations, but also on how agents interact and evolve across rounds. We propose SAUCE (Sequential Agent Uncertainty through Consensus Evolution), a lightweight, training-free uncertainty estimator that formulates MAS uncertainty as sequential inference over a latent system-level belief. SAUCE aggregates round-level agreement and generation-uncertainty signals through a filtering-style update. Across five backbones, five benchmarks, and two MAS protocols, SAUCE improves misclassification detection, selective prediction, and calibration over a broad set of uncertainty estimation baselines, including standard log-likelihood-based methods and MAS-specific estimators.

Problem

• Uncertainty Estimation
• Multi-Agent Systems
• Parallel Reasoning
• Large Language Models

Innovation

• Uncertainty Estimation
• Multi-Agent Systems
• Sequential Inference
• Consensus Evolution
• Training-Free