NSF-SciFy: Mining the NSF Awards Database for Scientific Claims

Researchers released NSF-SciFy, a large dataset mined from 400,000 NSF award abstracts.

The dataset contains 2.8 million claims, plus smaller focused subsets for targeted use.

Fine-tuning models on NSF-SciFy significantly improved scientific claim and proposal extraction.

The resource is designed to support claim verification, discovery tracking, and meta-science research across fields.