EUVIMEDOpdag EUVIMED
VergleichsstudieOpen access

Fine-Tuning Methods for Large Language Models in Clinical Medicine by Supervised Fine-Tuning and Direct Preference Optimization: Comparative Evaluation

Thomas Savage, Stephen P Ma, Abdessalem Boukil, Ekanath Rangan, Vishwesh Patel, Ivan Lopez, Jonathan Chen

Journal of Medical Internet Research · 2025-09-23 · 27 · e76048-e76048

DOI: 10.2196/76048 ↗

At a glance

Vergleichsstudie in Journal of Medical Internet Research, published online on 2025-09-23. Written by Thomas Savage and 6 co-authors. Institution: Division of Hospital Medicine, Perelman School of Medicine, Department of Medicine, University of Pennsylvania, 3400 Spruce St, Philadelphia, PA, 191…. Major MEDLINE topics: Natural Language Processing, Clinical Medicine. Cited 23 times so far (Crossref); 36 references. The full text is freely available in PubMed Central. Funded by: NIAID NIH HHS, NINDS NIH HHS, NIDA NIH HHS.

AbstractSubjectsBibliografiske oplysningerCiteRelated studies

Abstract (original language)

Background

Large language model (LLM) fine-tuning is the process of adjusting out-of-the-box model weights using a dataset of interest. Fine-tuning can be a powerful technique to improve model performance in fields like medicine, where LLMs may have poor out-of-the-box performance. The 2 common fine-tuning techniques are supervised fine-tuning (SFT) and direct preference optimization (DPO); however, little guidance is available for when to apply either method within clinical medicine or health care operations.

Objective

This study aims to investigate the benefits of fine-tuning with SFT and DPO across a range of core natural language tasks in medicine to better inform clinical informaticists when either technique should be deployed.

Methods

We use Llama3 8B (Meta) and Mistral 7B v2 (Mistral AI) to compare the performance of SFT alone and DPO across 4 common natural language tasks in medicine. The tasks we evaluate include text classification, clinical reasoning, text summarization, and clinical triage.

Results

Our results found clinical reasoning accuracy increased from 7% to 22% with base Llama3 and Mistral2, respectively, to 28% and 33% with SFT, and then 36% and 40% with DPO ( P =.003 and P =.004, respectively). Summarization quality, graded on a 5-point Likert scale, was 4.11 with base Llama3 and 3.93 with base Mistral2. Performance increased to 4.21 and 3.98 with SFT and then 4.34 and 4.08 with DPO ( P

Subjects

MeSH terms (MEDLINE)

Natural Language ProcessingClinical MedicineHumansLanguageLarge Language Models

Author keywords

artificial intelligencedirect preference optimizationfine-tuninglarge language modelssupervised fine-tuning

Authors and institutions

  • Division of Hospital Medicine, Perelman School of Medicine, Department of Medicine, University of Pennsylvania, 3400 Spruce St, Philadelphia, PA, 19147, United States, 1 2155191670.
  • Division of Hospital Medicine, Department of Medicine, Stanford Medicine, Palo Alto, CA, United States.
  • Linguamind AI, Sousse, Tunisia.
  • Department of Medicine, Stanford Medicine, Palo Alto, CA, United States.
  • Department of Medicine, Saint Michael's Medical Center, Newark, New Jersey, United States.
  • Center for Biomedical Informatics Research, Stanford University, Palo Alto, CA, United States.

Funding

  • NIAID NIH HHS
  • NINDS NIH HHS
  • NIDA NIH HHS
  • NCATS NIH HHS

Cite

APA

Savage, T., P Ma, S., Boukil, A., Rangan, E., Patel, V., Lopez, I., & Chen, J. (2025). Fine-Tuning Methods for Large Language Models in Clinical Medicine by Supervised Fine-Tuning and Direct Preference Optimization: Comparative Evaluation. Journal of Medical Internet Research, 27, e76048-e76048. https://doi.org/10.2196/76048

Vancouver

Savage T, P Ma S, Boukil A, Rangan E, Patel V, Lopez I, et al. Fine-Tuning Methods for Large Language Models in Clinical Medicine by Supervised Fine-Tuning and Direct Preference Optimization: Comparative Evaluation. Journal of Medical Internet Research. 2025;27:e76048-e76048. doi:10.2196/76048

BibTeX
@article{Savage2025,
  author = {Savage, Thomas and P Ma, Stephen and Boukil, Abdessalem and Rangan, Ekanath and Patel, Vishwesh and Lopez, Ivan and Chen, Jonathan},
  title = {Fine-Tuning Methods for Large Language Models in Clinical Medicine by Supervised Fine-Tuning and Direct Preference Optimization: Comparative Evaluation},
  journal = {Journal of Medical Internet Research},
  year = {2025},
  volume = {27},
  pages = {e76048--e76048},
  doi = {10.2196/76048}
}
Start for free →