hexuan21/VideoFeedback_r2k_bad8k

该数据集包含两个配置：annotated和real。每个配置都包含以下特征：id（字符串类型）、images（字符串序列）、text prompt（字符串类型）、video link（字符串类型）、visual quality（int64类型）、temporal consistency（int64类型）、dynamic degree（int64类型）、text-to-video alignment（int64类型）、factual consistency（int64类型）和conversations（包含from和value两个字段的列表）。数据分割仅包含训练集，annotated配置的训练集包含40,101个示例，real配置的训练集包含2,800个示例。The dataset includes two configurations (annotated and real), each containing multiple features such as id, image sequences, text prompts, video links, visual quality, temporal consistency, dynamic degree, and text-to-video alignment. The dataset is used for training and involves multimodal analysis of videos and text, with evaluation dimensions including visual quality, temporal consistency, dynamic degree, and text-to-video alignment.