Stanford NLI Corpus (SNLI)

Tags: nli, datasets

• 550k examples
• big enough to use NN’s without pre-training
• caption-based