MahmoodAnaam/raid-dataset
收藏资源简介：
该数据集是一个对抗性攻击文本数据集，包含文本样本及其相关元数据，用于评估模型在对抗性攻击下的性能。特征包括唯一标识符（id）、对抗源标识符（adv_source_id）、源标识符（source_id）、文本内容（text）、模型信息（model）、攻击类型（attack）、体裁（genre）和二分类标签（label，0或1）。数据集分为开发集（dev）和测试集（test），分别包含559,860和373,240个样本，总大小约1.55 GB。
This dataset is an adversarial attack text dataset containing text samples and related metadata for evaluating model performance under adversarial attacks. Features include unique identifier (id), adversarial source identifier (adv_source_id), source identifier (source_id), text content (text), model information (model), attack type (attack), genre (genre), and binary classification label (label, 0 or 1). The dataset is split into development (dev) and test sets, with 559,860 and 373,240 samples respectively, and a total size of approximately 1.55 GB.
