Accessibility settings

This paper is in the following e-collection/theme issue:

Artificial Intelligence (2501) Clinical Information and Decision Making (2178) Research Instruments, Questionnaires, and Tools (813) Decision Support for Health Professionals (1724) Generative Language Models Including ChatGPT (913) AI Language Models in Health Care (305) Responsible Health AI (115)

Published on in Vol 28 (2026)

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/90061, first published .
Person using a laptop with ChatGPT medical question interface open

Benchmark Integrity and Reasoning-Trace Errors in Medical Question Answering With Large Language Models: Mixed Methods Study With Sparse Autoencoders

Benchmark Integrity and Reasoning-Trace Errors in Medical Question Answering With Large Language Models: Mixed Methods Study With Sparse Autoencoders

Authors of this article:

Jialin Liu1, 2 Author Orcid Image ;   Siru Liu3, 4 Author Orcid Image ;   Adam Wright3, 5 Author Orcid Image
Article Authors Cited by Tweetations Metrics

Jialin Liu   1, 2 * , MD ;   Siru Liu   3, 4 * , PhD ;   Adam Wright   3, 5 , PhD

1 Department of Medical Informatics, West China Hospital of Sichuan University, Chengdu, Sichuan, China

2 Department of Otolaryngology-Head and Neck Surgery, West China Hospital of Sichuan University, Chengdu, Sichuan, China

3 Department of Biomedical Informatics, Vanderbilt University Medical Center, Nashville, TN, United States

4 Department of Computer Science, Vanderbilt University, Nashville, TN, United States

5 Department of Medicine, Vanderbilt University Medical Center, Nashville, TN, United States

*these authors contributed equally

Corresponding Author:

  • Siru Liu, PhD
  • Department of Biomedical Informatics
  • Vanderbilt University Medical Center
  • 2525 West End Ave
  • Nashville, TN 37203
  • United States
  • Phone: 1 615 936 6867
  • Email: siru.liu@vumc.org