Community
2025/10/18
Specialized vs. General-Purpose LLMs — and How to Compare Them
A recent Nature study introduced Delphi-2M, a specialized medical LLM trained on 2.3 million individuals’ EHRs to predict over 1,000 diseases, achieving an average AUC of 0.67. While impressive, it raises key questions: Can general-purpose LLMs like ChatGPT or Gemini outperform specialized models in disease prediction? And how should such models be benchmarked in realistic clinical settings? Initiatives like Stanford’s MedAgentBench and the JAMA Summit are beginning to address these challenges, shaping the future of evaluating LLMs for real-world healthcare.
2025/6/1
We are excited to share that ELHS Institute’s GenAI-ELHS Solution has been selected as a winner in the recent global competition, Transforming Cancer Navigation with Open Data & APIs Challenge, sponsored by the Robert Wood Johnson Foundation and AcademyHealth. We are proud that our innovations in democratizing generative AI (GenAI) and Learning Health Systems (LHS) to transform predictive health care have been recognized by leading organizations in healthcare.
2024/9/29
The first in a series of free practical training webinars designed to teach doctors, medical students, and healthcare professionals how to quickly start using their own LLM copilots for healthcare. [In Chinese]
