Reproduce “Towards Accurate Differential Diagnosis with Large Language Models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Daniel McDuff, Mike Schaekermann, Tao Tu, Anil Palepu, Amy Wang, Jake Garrison, Karan Singhal, Yash Sharma and 20 more
Researchers at Google developed and evaluated a specialized Large Language Model (LLM) designed for differential diagnosis using challenging clinical cases from the New England Journal of Medicine. The LLM demonstrated higher standalone diagnostic accuracy than unassisted clinicians and, when used as an assistive tool, significantly enhanced clinicians' ability to generate more accurate, comprehensive, and appropriate differential diagnoses.
Key insights
- Specialized Large Language Models (LLMs) can achieve high standalone performance in complex differential diagnosis, outperforming unassisted human clinicians on challenging cases.
- LLMs function effectively as assistive tools, significantly improving the diagnostic accuracy, comprehensiveness, and appropriateness of human clinicians' differential lists.
- Clinicians found the LLM interface intuitive, efficient, and particularly helpful when initial differentials were unclear, suggesting its potential roles in medical education and clinical support.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.