shalanova/benchmark-1-arabic-gt

该数据集主要包含提示注入和典型的越狱风格指令，攻击模式相对单一。数据集大小为1,000个提示（500个安全/500个不安全），包含四个列：text（原始提示）、label（0表示安全，1表示不安全）、translation（通过Google Translate翻译成阿拉伯语的提示）和score_ar_google（与codebook的余弦相似度得分）。数据集来源于jayavibhav/prompt-injection-safety，更多信息可参考提供的论文链接。

This dataset primarily contains prompt-injection and canonical jailbreak-style instructions with relatively homogeneous attack patterns. The dataset size is 1,000 prompts (500 safe / 500 unsafe), with four columns: text (original prompt), label (0: safe, 1: unsafe), translation (prompt translated into Arabic by Google Translate), and score_ar_google (cosine similarity score with codebook). The dataset is sourced from jayavibhav/prompt-injection-safety, and more information can be found in the provided paper link.