Multiple Choice Questions and Large Languages Models: A Case Study with Fictional Medical Data