Are LLM Belief Updates Consistent with Bayes’ Theorem?
Published in Under review @AAAI, Accepted @ICML 2025 Workshop on Assessing World Models, 2025
This paper evaluates whether large language models (LLMs) update their beliefs over propositions in a Bayesian-consistent manner as they scale in size and capability.
Recommended citation: Sohaib Imran, Ihor Kendiukhov, Matthew Broerman, Aditya Thomas, Riccardo Campanella, Rob Lamb, Peter M. Atkinson.
Download Paper | Download Slides