BayesAgent: Bayesian Agentic Reasoning Under Uncertainty via Verbalized Probabilistic Graphical Modeling

TL;DR

This paper introduces vPGM, a Bayesian framework that integrates probabilistic graphical models with Large Language Models to improve reasoning and uncertainty modeling without requiring expert-designed models.

Contribution

The paper presents vPGM, a novel verbalized probabilistic graphical modeling approach that guides LLMs in agentic reasoning under uncertainty without expert-driven model design.

Findings

Enhanced confidence calibration in LLMs.

Improved text generation quality.

Effective reasoning in agentic tasks.

Abstract

Human cognition excels at transcending sensory input and forming latent representations that structure our understanding of the world. While Large Language Model (LLM) agents demonstrate emergent reasoning and decision-making abilities, they lack a principled framework for capturing latent structures and modeling uncertainty. In this work, we explore for the first time how to bridge LLM agents with probabilistic graphical models (PGMs) to address agentic reasoning under uncertainty. To this end, we introduce Verbalized Probabilistic Graphical Modeling (vPGM), a Bayesian agentic framework that (i) guides LLM agents in following key principles of PGMs through natural language and (ii) refines the resulting posterior distributions via numerical Bayesian inference. Unlike many traditional probabilistic methods requiring substantial domain expertise, vPGM bypasses expert-driven model design, making it well-suited for scenarios with limited assumptions. We evaluated our model on several agentic reasoning tasks, both close-ended and open-ended. Our results indicate that the model effectively enhances confidence calibration and text generation quality.