MSTS
收藏资源简介:
MSTS Benchmark数据集是一个用于多模态安全测试的基准数据集,包含多种语言的数据,如德语、俄语、中文、印地语、西班牙语、意大利语、法语、英语、韩语、阿拉伯语和波斯语。数据集的特征包括危险类别、子类别、子子类别、案例ID、案例文本、不安全图像ID、不安全图像描述、提示文本、提示类型、不安全图像URL、不安全图像许可证、不安全图像内容警告以及不安全图像本身。数据集的分割部分显示了每种语言的字节数和示例数。数据集的许可证为cc-by-4.0,任务类别为图像文本到文本,标签为不适合所有观众。数据集的使用示例和引用信息也被提供。
The MSTS Benchmark dataset is a benchmark dataset for multimodal safety testing, containing data in multiple languages including German, Russian, Chinese, Hindi, Spanish, Italian, French, English, Korean, Arabic and Persian. The dataset features hazard category, subcategory, sub-subcategory, case ID, case text, unsafe image ID, unsafe image description, prompt text, prompt type, unsafe image URL, unsafe image license, unsafe image content warning, and the unsafe images themselves. The dataset split section displays the byte count and sample count for each language. The dataset is licensed under cc-by-4.0, with its task category being image-text to text and the label being "Not suitable for all audiences". Usage examples and citation information of the dataset are also provided.




