prli/pubmed-full_draft_chopped-rare-qwen-alpha0-80_perturb

该数据集包含原始文本数据及其经过多种扰动处理后的变体版本。主要特征包括：原始文本(text)和六种文本扰动类型(同义词替换、butter_fingers扰动、随机删除、字符大小写变化、空格扰动和下划线技巧)。数据集仅包含验证集，共4000个样本。

This dataset contains original text data and its variants processed through various perturbations. Key features include: original text and six types of text perturbations (synonym substitution, butter fingers, random deletion, character case change, whitespace perturbation, and underscore trick). The dataset only includes a validation set with 4000 examples.