notesum.ai
Published at November 13The Limited Impact of Medical Adaptation of Large Language and Vision-Language Models
cs.CL
cs.AI
cs.LG
Released Date: November 13, 2024
Authors: Daniel P. Jeong1, Pranav Mani2, Saurabh Garg3, Zachary C. Lipton1, Michael Oberst4
Aff.: 1Machine Learning Department, Carnegie Mellon University; 2Abridge AI; 3Unknown Affiliation; 4Department of Computer Science, Johns Hopkins University

| Model Class | General Domain | Medical Domain | Medical Adaptation Corpora |
|---|---|---|---|
| LLM | Llama-3-70B-Instruct (Meta, 2024) | Med42-v2-70B (Christophe et al., 2024b) | Medical QA Datasets (e.g., MedQA, MedMCQA) |
| Medical Instruction 120k (Altaf, 2023) | |||
| OpenGPT (OpenChat) (Wang et al., 2024) | |||
| StackExchange (Lambert et al., 2023) | |||
| Medical Flashcards (Han et al., 2023) | |||
| Llama-3-70B-Instruct (Meta, 2024) | OpenBioLLM-70B (Pal and Sankarasubbu, 2024) | Undisclosed | |
| Llama-2-70B (Touvron et al., 2023b) | MediTron-70B (Chen et al., 2023) | Clinical Practice Guidelines (e.g., CDC, WHO) | |
| PubMed Articles (S2ORC; Lo et al., 2020) | |||
| Llama-2-70B (Touvron et al., 2023b) | Clinical-Camel-70B (Toma et al., 2023) | ShareGPT | |
| 20k PubMed Articles Published Before 2021 | |||
| Random 4k Subset of MedQA (Jin et al., 2020) | |||
| Llama-2-70B (Touvron et al., 2023b) | Med42-v1-70B (Christophe et al., 2024a) | Medical QA Datasets (e.g., MedQA, MedMCQA) | |
| OpenGPT (OpenChat) (Wang et al., 2024) | |||
| StackExchange (Lambert et al., 2023) | |||
| Medical Flashcards (Han et al., 2023) | |||
| CORD-19 (Wang et al., 2020) | |||
| Llama-3-8B-Instruct (Meta, 2024) | Med42-v2-8B (Christophe et al., 2024b) | Medical QA Datasets (e.g., MedQA, MedMCQA) | |
| Medical Instruction 120k (Altaf, 2023) | |||
| OpenGPT (OpenChat) (Wang et al., 2024) | |||
| StackExchange (Lambert et al., 2023) | |||
| Medical Flashcards (Han et al., 2023) | |||
| Llama-3-8B (Meta, 2024) | OpenBioLLM-8B (Pal and Sankarasubbu, 2024) | Undisclosed | |
| Llama-2-7B (Touvron et al., 2023b) | MediTron-7B (Chen et al., 2023) | Clinical Practice Guidelines (e.g., CDC, WHO) | |
| PubMed Articles (S2ORC; Lo et al., 2020) | |||
| Mistral-7B-Instruct-v0.1 (Jiang et al., 2023) | BioMistral-7B (Labrak et al., 2024) | PubMed Articles (PMC Open Access Subset) | |
| Llama-2-7B-Chat (Touvron et al., 2023b) | BioMedGPT-LM-7B (Luo et al., 2023) | PubMed Articles (S2ORC; Lo et al., 2020) | |
| VLM | LLaVA-v0-7B (Liu et al., 2023) | LLaVA-Med-7B (Li et al., 2023a) | PubMed Articles (PMC-15M; Zhang et al., 2023) |
| Open-Flamingo-9B (Awadalla et al., 2023) | Med-Flamingo-9B (Moor et al., 2023b) | Medical Textbooks (MTB; Moor et al., 2023b) | |
| PubMed Articles (PMC-OA; Lin et al., 2023) |