mmfm-trust/V2T

该数据集包含五个部分：1) adv部分包含问题、答案和对抗性示例等信息，适用于回答相关任务；2) fairness部分包含视频和与公平性评估相关的信息，适用于公平性研究；3) hallucination部分包含场景、任务和问题回答等信息，适用于幻觉生成任务；4) privacy部分包含文件和位置信息，适用于隐私研究；5) safety部分包含视频内容和类别信息，适用于安全性研究。

The dataset consists of five parts: 1) The adv part contains information on questions, answers, and adversarial examples, suitable for answer-related tasks; 2) The fairness part contains videos and information related to fairness evaluation, suitable for fairness research; 3) The hallucination part contains information on scenarios, tasks, and question answering, suitable for hallucination generation tasks; 4) The privacy part contains file and location information, suitable for privacy research; 5) The safety part contains video content and category information, suitable for safety research.