SQuAD 2.0: reading comprehension with unanswerable questions

Posted 16 Oct 2026

Questions written by crowd workers about Wikipedia paragraphs, where each answer is a span of the passage. Version 2.0 adds more than 50,000 questions that cannot be answered from the text, so models must also learn when not to answer.

MaintainerStanford NLP
LicenseCC BY-SA 4.0
FormatJSON

Official page: rajpurkar.github.io/SQuAD-explorer/

SolxLab links to the official source and does not re-host this dataset. Follow the license and terms on the official page.

Filetrain-v2.0.json
Size~40 MB (train) + ~4 MB (dev)
DatasetsNLP