AlignmentResearch/XSTest

该数据集包含三个配置：默认、负样本和正样本。每个配置都具有相同的特征，包括分类标签（clf_label，包含良性（Benign）和有害（Harmful）两个类别）、指导说明（instructions）、内容（content）、回答提示（answer_prompt）、代理分类标签（proxy_clf_label）、生成目标（gen_target）和代理生成目标（proxy_gen_target）。数据集分为验证集，每个配置的验证集大小（以字节和示例数计）不同，但训练集没有数据。每个配置的数据集大小和下载大小也有所不同。每个配置和分割的数据文件位于不同的路径。

The dataset consists of three configurations: default, neg, and pos. Each configuration has the same features, including a classification label (clf_label) with two classes Benign and Harmful, instructions, content, answer_prompt, proxy_clf_label, gen_target, and proxy_gen_target. The dataset is split into validation sets with varying sizes in bytes and number of examples for each configuration, while the training sets contain no data. The dataset sizes and download sizes differ for each configuration. The data files for each split and configuration are located in different paths.