charisgao/human-raid-clean

该数据集包含749,970个样本，主要用于训练目的。每个样本包含多个特征字段，如id、adv_source_id、source_id、model、decoding、repetition_penalty、attack、domain、title、prompt和generation等。这些字段描述了样本的不同属性，如模型类型、解码参数、重复惩罚、攻击类型、领域、标题、提示和生成内容等。数据集的总大小为983,144,706字节，下载大小为587,787,853字节。

The dataset contains 749,970 samples, primarily intended for training purposes. Each sample includes multiple feature fields such as id, adv_source_id, source_id, model, decoding, repetition_penalty, attack, domain, title, prompt, and generation. These fields describe various attributes of the samples, such as model type, decoding parameters, repetition penalty, attack type, domain, title, prompt, and generated content. The total size of the dataset is 983,144,706 bytes, with a download size of 587,787,853 bytes.