SNLI 1.0 Dataset

570,152 train/dev/test sentence pairs. The public preview shows verified label counts without publishing original sentences. The private ZIP contains the source JSONL splits, source README, attribution, and CC BY-SA 4.0 notice. Full download requires a paid Dataset application; target-language translations are being refreshed.

Dataset

SNLI

SNLI 1.0: 570,152 train/dev/test sentence pairs in a private ZIP with the official JSONL splits, source README, attribution and CC BY-SA 4.0 notice. Public preview shows aggregate labels only; the full sentence text is available after purchase.

570,152Records

47.2 MBFile size

ZIP / JSONLDelivery formats

Data sample

Field reference

The purchased version's schema snapshot is authoritative.

Data quality

Quality metrics and processing details come from this version's acceptance report.

dev_pairs10000

test_pairs10000

train_pairs550152

unlabeled_pairs1119

License

Stanford NLP SNLI 1.0 under CC BY-SA 4.0; source README includes Flickr30k/VisualGenome provenance and citation obligations.

SNLI Stanford Natural Language Inference Dataset