A study titled 'Re-evaluating GPT-4’s bar exam performance' critically examines the claim of GPT-4's 90th percentile score on the Uniform Bar Exam, revealing methodological shortcomings and suggesting overinflated estimates of its legal reasoning capabilities.