prli/uspto-full_draft_chopped-rare-qwen-alpha0-20_perturb

该数据集包含多种文本扰动特征，如同义词替换、butter fingers、随机删除、字符大小写变化、空格扰动和下划线技巧。数据集仅包含一个名为validation的分割，包含4,000个样本，总大小为101,489,798字节。该数据集设计用于验证目的，如分割名称所示，没有训练或测试分割。

The dataset includes various text perturbation features such as synonym substitution, butter fingers, random deletion, change char case, whitespace perturbation, and underscore trick. It has a single split named validation with 4,000 examples and a total size of 101,489,798 bytes. The dataset is designed for validation purposes, as indicated by the split name and the absence of training or test splits.