train: 6,000 examples are randomly sampled from "train" split of SNLI dataset out of which 3000 examples are left unmodified and universal triggers are prepended to the hypothesis of rest 3000 examples.
entailment: ~1000 randomly chosen entailment examples from validation split of SNLI dataset.
entailment_triggers: ~1000 examples of entailment split of this dataset are prepended with universal… See the full description on the dataset page:
https://huggingface.co/datasets/ckverma/snli_universal.
