Rapidata/text-2-video-human-preferences-veo3

Pika 2.2人类偏好数据集，包含约4.6万份来自约2万名标注者的反馈，用于评估Veo3视频生成模型的表现。数据集包括对齐度、连贯性和偏好评分，以及详细结果。支持视频分类、文本到视频和文本分类任务，使用Apache-2.0许可证。

Pika 2.2 Human Preferences dataset, containing about 46,000 responses from approximately 20,000 annotators, used to evaluate the performance of the Veo3 video generation model. The dataset includes alignment, coherence, and preference scores, as well as detailed results. It supports video classification, text-to-video, and text classification tasks, and is licensed under Apache-2.0.