Solving Bayesian reasoning tasks with ChatGPT and Gemini

Abstract

This paper expands on the exploration of Bayesian problem-solving capabilities in large language models (LLMs), specifically ChatGPT, and Gemini. Building upon our prior study, where ChatGPT excelled in solving 10 Bayesian problems, we extend the scope by introducing four additional tasks to both ChatGPT and Gemini in order to compare performance. The results demonstrate ChatGPT and Gemini consistent accuracy in tackling all four reasoning problems presented. The obtained results suggest the potential of LLM like ChatGPT and Gemini for effectively handling Bayesian reasoning tasks relevant to science and engineering fields.