Skip to content

Dataset · mteb

Scientific claim verification benchmark based on article abstracts

mteb/scifact

English dataset from the MTEB suite for evaluating evidence retrieval in scientific publications and medical texts.

What's inside

Part of the well-known MTEB embedding benchmark, originally created at AllenAI. Contains between 1 and 10 thousand English records for the text-retrieval task: the model must find supporting or refuting evidence for given scientific claims within a database of paper abstracts. Covers academic and medical domains.

What you can do with it

  • Evaluating text embedding quality and retrievers for scientific and medical RAG
  • Comparing search models on the standardized MTEB benchmark
Tasks
Text retrieval
Languages
English
Size
1K–10K rows
Downloads this month
40.1K
Likes
5

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("mteb/scifact")