Comparative Performance of Large Language Models in Answering Guideline-Based Questions on Juvenile Idiopathic Arthritis: A Blinded Expert Evaluation
In a blinded expert evaluation of guideline-based juvenile idiopathic arthritis questions, GPT-5.6 Sol significantly outperformed Gemini 3.1 Pro and Claude Opus 5 in accuracy, usability, and overall quality, supporting the need for supervised rather than autonomous use of large language models in pediatric rheumatology.