prli/uspto-full_draft_chopped-rare-qwen-alpha0-50_perturb

该数据集包含文本数据及其经过多种扰动处理的版本，具体特征包括原始文本、同义词替换、butter_fingers（模拟打字错误）、随机删除、字符大小写变化、空格扰动和下划线技巧。数据集可能用于自然语言处理任务，如文本增强、模型鲁棒性评估或数据扩增实验，仅包含验证集，共4000个示例。

This dataset contains text data along with multiple perturbed versions, including features such as original text, synonym substitution, butter_fingers (simulating typing errors), random deletion, change in character case, whitespace perturbation, and underscore trick. It is likely intended for natural language processing tasks, such as text augmentation, model robustness evaluation, or data augmentation experiments, and includes only a validation split with 4000 examples.