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

Published: 2 May 2025| Version 1 | DOI: 10.17632/79tbbm7v24.1
Contributor:
Fabio Borgonovo

Description

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.

Files

Institutions

  • Mayo Clinic Rochester

Categories

Large Language Model

Licence