PKU-Alignment/MM-SafetyBench

MM-SafetyBench 数据集是一个包含多个子数据集的综合数据集，每个子数据集针对不同的应用场景，如经济伤害、金融建议、欺诈等。每个子数据集都包含问题、图片和相关标识符等信息，并且根据不同的数据类型和错误类型（如拼写错误）进行划分。数据集适用于研究用途，受特定的非商业许可协议约束。

MM-SafetyBench dataset is a comprehensive dataset containing multiple sub-datasets, each targeting different application scenarios such as economic harm, financial advice, fraud, etc. Each sub-dataset includes questions, images, and related identifiers, and is divided according to different data types and error types (such as spelling mistakes). The dataset is intended for research purposes and is subject to a specific non-commercial license agreement.