The first publicly shareable clinical LLM (Edward Choi's lab): trains on synthetic clinical notes generated from public PMC case reports, sidestepping the privacy restrictions that keep real-EHR models unshareable, and released Llama-family 7B/13B checkpoints that match models trained on real notes. Anchor of the lab's medical-FM line that continued to LabTOP (CHIL 2025 Best Paper, generative lab-test-outcome prediction on EHR) and the CXReasonBench and PatientSim NeurIPS 2025 D&B Spotlights.

Paper

medicalopen-weight