AI Research Atlas

Med-PaLM 2

Google · 16 May 2023

Med-PaLM 2 scored 86.5% on MedQA (USMLE-style), 19+ points above Med-PaLM, with physicians preferring its answers on 8 of 9 axes.

Medical fine-tune of PaLM 2 using an 'ensemble refinement' prompting method. Physicians rated its long-form answers better than physician-written ones on 8 of 9 clinical-utility axes. Authors stress it is not clinically validated.

Date
Tuesday, 16 May 2023
Lab
Google
Kind
model
Access
research preview

Figures

MeasureValueMeasured by
MedQA accuracy86.5%
vs 67.2% for Med-PaLM
company

arXiv v1 2023-05-16; the PaLM 2 blog (2023-05-10) already listed Med-PaLM 2. Limited Google Cloud access for customers.

Sources

  1. arxiv.org/abs/2305.09617
  2. blog.google/technology/ai/google-palm-2-ai-large-language-model/

This record was checked against its sources on 6 October 2026. How we check

Related