# MedAlpaca -- An Open-Source Collection of Medical Conversational AI Models and Training Data

**Authors:** Tianyu Han, Lisa C. Adams, Jens-Michalis Papaioannou, Paul Grundmann, Tom Oberhauser, Alexander Löser, Daniel Truhn, Keno K. Bressem  
**arXiv:** [2304.08247](https://arxiv.org/abs/2304.08247)  
**Date:** 2023-04-14  
**Summary page:** https://www.summarizepaper.com/en/paper/2304.08247v1/

## Key points

- Large language models (LLMs) like OpenAI's GPT series have potential in various fields
- LLMs can enhance medical workflows, diagnostics, patient care, and education
- Concern for patient privacy protection necessitates open-source models that can be deployed on-premises
- Researchers introduce a dataset for fine-tuning LLMs in medical contexts
- Fine-tuning improves the accuracy and effectiveness of LLMs for medical certifications
- Utilizing open-source models and fine-tuning techniques can enhance medical practices while safeguarding patient privacy.

## Summary

In their research, Tianyu Han, Lisa C. Adams, Jens-Michalis Papaioannou, Paul Grundmann, Tom Oberhauser, Alexander Löser, Daniel Truhn and Keno K. Bressem highlight the potential of large language models (LLMs) like OpenAI's GPT series in various fields due to their artificial intelligence capabilities. Specifically focusing on the medical field, the authors emphasize how LLMs can greatly enhance medical workflows, diagnostics, patient care and education. However, one major concern is the need for open-source models that can be deployed on-premises to ensure patient privacy protection. To address this issue and enable effective use of LLMs in medical applications, the researchers introduce an innovative dataset comprising over 160 000 entries. This dataset is specifically designed for fine-tuning LLMs to optimize their performance in medical contexts. The study investigates the impact of fine-tuning these datasets on publicly accessible pre-trained LLMs. The authors compare the performance of pre-trained only models with that of fine-tuned models in relation to the examinations required for medical certification. By doing so they aim to demonstrate how fine-tuning can significantly improve the accuracy and effectiveness of LLMs in meeting the requirements for future medical doctors' certifications. Overall this research highlights the potential benefits of utilizing open source models and fine tuning techniques to harness the power of LLMs in enhancing medical practices while safeguarding patient privacy.