Lidor-Mashiach/mnli-training-paraphrase-augmentation

MNLI Training Paraphrase Augmentation Purpose This dataset contains new paraphrases created for training paraphrase augmentation during Phase B of the research project. It was not used as an NLI evaluation set. It was not used for paraphrase consistency evaluation. It is separate from the published MNLI Paraphrase Bank used for evaluation. The generation records report zero collisions with that evaluation bank. The CSV contains only the new augmentation rows. It… See the full description on the dataset page: https://huggingface.co/datasets/Lidor-Mashiach/mnli-training-paraphrase-augmentation.