Teaching LLMs to reason like Bayesians

Google research explores methods to teach large language models how to reason using Bayesian principles, improving their analytical capabilities.