Skip to content

Dataset · MedRAG

PubMed corpus for medical RAG

MedRAG/pubmed

English PubMed text snippets for medical Retrieval-Augmented Generation.

What's inside

Contains PubMed corpus snippets used in MedRAG. Designed for medical RAG; English language, text modality. Metadata size is 1M–10M records. The card notes updated id and PMID columns.

What you can do with it

  • Building and evaluating medical RAG
  • Training retriever/embedder for biomedical texts
  • Testing QA systems on PubMed
Tasks
Question answering
Languages
English
Size
1M–10M rows
Downloads this month
11.2K
Likes
117

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("MedRAG/pubmed")