Related Experiment Video
Updated: May 16, 2025
03:14
Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
Published on: December 6, 2024
466
The Development Landscape of Large Language Models for Biomedical Applications
Zhiyuan Cao1, Vipina K Keloth1, Qianqian Xie1
1Department of Biomedical Informatics and Data Science, School of Medicine, Yale University, New Haven, Connecticut, USA;
Annual Review of Biomedical Data Science
|April 1, 2025
Summary
This review explores large language models (LLMs) in biomedicine, finding decoder-only architectures and biomedical literature are common. Challenges include data privacy and model transparency.
Area of Science:
- Biomedical Informatics
- Artificial Intelligence in Healthcare
Background:
- Large language models (LLMs) show significant promise for transforming healthcare and medical research.
- The release of ChatGPT in 2022 spurred rapid development and application of LLMs in diverse biomedical fields.
Purpose of the Study:
- To review the development landscape of text-based biomedical LLMs.
- To analyze their characteristics, development processes, and applications.
Main Methods:
- Systematic review adhering to PRISMA guidelines.
- Selection of 82 articles from 5,512 published since 2022, focusing on LLMs trained with biomedical data.
Main Results:
- Predominant use of decoder-only architectures (e.g., Llama 7B).
- Prevalence of task-specific fine-tuning strategies.
- Heavy reliance on biomedical literature for model training.
Conclusions:
- Ongoing challenges include balancing data openness with privacy and detailing computational resources.
- Future directions involve multimodal integration, specialized medical LLMs, and enhanced data/model accessibility.
