Battle of the Bots: Solving Clinical Cases in Osteoarticular Infections with Large Language Models

This repository contains the three core datasets underpinning our LLM evaluation study in infectious disease decision‐making: Multiple‐Choice Answers: Model responses mapped to predefined, guideline‐based answer keys for each clinical question. Raw LLM Outputs: Unedited textual answers generated by all 15 tested language models. Likert‐Scale Ratings: Explanation‐quality scores assigned by two blinded, board‐certified reviewers, including consensus‐resolved discrepancies and interrater reliability statistics. Together, these files enable full replication of our accuracy and explanation‐quality analyses.