prli/pubmed-full_draft_chopped-rare-qwen-alpha1-00_perturb

该数据集包含4000个验证示例，每个示例具有原始文本和多种扰动版本，如同义词替换、随机字符删除、字符大小写变化、空格扰动和下划线技巧等，用于文本增强或模型鲁棒性评估。特征包括text（原始文本）、synonym_substitution（同义词替换）、butter_fingers（打字错误模拟）、random_deletion（随机删除）、change_char_case（大小写变化）、whitespace_perturbation（空格扰动）和underscore_trick（下划线技巧）。数据集仅提供验证分割，下载大小约为23.7MB，总大小约为37.4MB。

This dataset contains 4000 validation examples, each with original text and multiple perturbed versions such as synonym substitution, random character deletion, character case changes, whitespace perturbation, and underscore tricks, for text augmentation or model robustness evaluation. Features include text (original), synonym_substitution, butter_fingers, random_deletion, change_char_case, whitespace_perturbation, and underscore_trick. The dataset provides only a validation split, with a download size of approximately 23.7MB and a total size of approximately 37.4MB.