README.md

metadata

source_datasets:
  - multi_nli
task_categories:
  - text-classification
task_ids:
  - text-scoring
  - semantic-similarity-scoring

Dataset Description

This dataset provides easier accessibility to the original MNLI dataset.

We randomly choose 10% of the original validation_matched split and use it as the validation split. The remaining 90% are used for the test split. The train split remains unchanged.