Dataset · allenai
Scientific claims with evidence from articles
allenai/scifact
Benchmark of 1.4 thousand scientific claims and abstracts for fact-checking and evidence retrieval.
What's inside
The dataset from AllenAI contains 1,400 expert-crafted scientific claims in English. Each claim is linked to scientific article abstracts containing evidence, along with veracity labels and textual rationales.
What you can do with it
- Evaluating the accuracy of RAG systems and fact-checking models in science
- Training classifiers for hypothesis verification on scientific texts
- Tasks
- Text classification
- Languages
- English
- Size
- 1K–10K rows
- Downloads this month
- 3.6K
- Likes
- 34
Models trained on it
- nikolamilosevic/SCIFACT_xlm_roberta_largeText classification 38.9K
- mradermacher/Aitana-7B-S-Instruct-i1-GGUFText generation 2.8K
- gplsi/Aitana-7B-S-InstructText generation 1.2K
- gplsi/Aitana-2B-S-InstructText generation 976
- mradermacher/Aitana-7B-S-Instruct-v0.1-i1-GGUFText generation 770
- mradermacher/Aitana-7B-S-Instruct-GGUFText generation 523
- mradermacher/Aitana-2B-S-Instruct-IP-1.0-GGUFText generation 340
- mradermacher/Aitana-2B-S-tourism-Instruct-1.0-GGUFText generation 293
How to load it
from datasets import load_dataset
ds = load_dataset("allenai/scifact")