Maintenance Notice

Due to necessary scheduled maintenance, the JMIR Publications website will be unavailable from Wednesday, July 01, 2020 at 8:00 PM to 10:00 PM EST. We apologize in advance for any inconvenience this may cause you.

Who will be affected?

Accepted for/Published in: JMIR AI

Date Submitted: Jul 22, 2025
Date Accepted: May 3, 2026

The final, peer-reviewed published version of this preprint can be found here:

Application of Sparse Autoencoders to Enhance Mechanistic Interpretability of Large Language Models in Medicine

Patil S, Metzger A, Karabacak M, Margetis K

Application of Sparse Autoencoders to Enhance Mechanistic Interpretability of Large Language Models in Medicine

JMIR AI 2026;5:e81134

DOI: 10.2196/81134

PMID: 42201744

Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.

Application of Sparse Autoencoders to Enhance Mechanistic Interpretability of Large Language Models in Medicine

  • Shiv Patil; 
  • Andre Metzger; 
  • Mert Karabacak; 
  • Konstantinos Margetis

ABSTRACT

Large language models (LLM) are positioned to transform the practice of medicine through their ability to determine clinical diagnoses, form treatment plans, and optimize medical workflows. Understanding the internal mechanism by which these models operate is necessary to legitimize LLMs in clinical practice. This field of research is called mechanistic interpretability. By extracting human-interpretable features from LLMs, sparse autoencoders (SAEs) represent a promising development in the field of digital medicine.


 Citation

Please cite as:

Patil S, Metzger A, Karabacak M, Margetis K

Application of Sparse Autoencoders to Enhance Mechanistic Interpretability of Large Language Models in Medicine

JMIR AI 2026;5:e81134

DOI: 10.2196/81134

PMID: 42201744

Download PDF


Request queued. Please wait while the file is being generated. It may take some time.

© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.