Brief

A new framework called Epistemic Regret Minimization (ERM) has been introduced to improve the causal reasoning of large language models. Unlike traditional methods that only reward correct answers, ERM critiques the underlying reasoning process itself. This label-free approach identifies and corrects issues like conflating correlation with causation and unexamined confounding variables within the model's thought process. Experiments show ERM significantly enhances the causal reasoning capabilities of models like GPT-4 Turbo and GPT-5.2, outperforming standard test-time correction methods. AI

A new study published on arXiv reveals that the language used to prompt large language models significantly impacts their diagnostic reasoning and accuracy in clinical settings. Researchers found that four out of five evaluated models performed better when prompted in English compared to French, with English yielding higher scores in differential diagnosis, logical structure, and internal validity. Only one model, o3, showed no significant language-based performance difference, highlighting the need to consider linguistic and cultural factors for equitable global deployment of LLMs in healthcare. AI