EPFL-Yale's Meditron: Open-Source Medical AI


Researchers from EPFL School of Computer and Communication Sciences and Yale School of Medicine have collaborated to create Meditron, a suite of open-source large multimodal foundation models tailored for the medical field. Leveraging Meta Llama 2, Meditron is designed to aid clinical decision-making and diagnosis, filling a crucial innovation gap in low-resource medical settings. Within the first months of its release, the model has been downloaded over 30,000 times, indicating its significant impact. Additionally, the team swiftly fine-tuned the 8-billion parameter model within 24 hours of Llama 3's release, resulting in Llama-3[8B]-MeditronV1.0. This version demonstrates robust performance on standard benchmarks like MedQA and MedMCQA, showcasing its potential for enhancing medical practices worldwide.