Inst-IT/Inst-It-Bench

Inst-It Bench是一个细粒度的多模态问答数据集，包含图像和视频两部分，每个部分都提供开放式问答和选择题两种形式。图像部分包含338张图像和1036个问答对，视频部分包含206个视频和1001个问答对。Inst-It Bench is a fine-grained multimodal QA dataset, which includes both image and video parts, each providing open-ended and multiple-choice formats. The image part contains 338 images and 1036 QA pairs, while the video part contains 206 videos and 1001 QA pairs.