MMInstruction/Video-T3-QA

文本时间理解数据集，用于将文本中的时间推理能力迁移到视频中。数据集提供了LLaVA风格的文本问答样本，包括视频关键帧描述和基于描述的问题。数据集包含顺序相关、属性相关、基于模板的顺序问题、时间指代问题以及时间定位问题等多种类型的问题。

Textual Temporal Understanding Dataset for transferring temporal reasoning ability from text to video. The dataset provides LLaVA-style text QA samples including video keyframe descriptions and questions based on them. It includes various types of questions such as order-related, attribute-related, template-based order questions, temporal referring questions, and temporal grounding questions.