Mitzi4132/VideoLLava
收藏Hugging Face2024-07-09 更新2024-07-22 收录
下载链接:
https://hf-mirror.com/datasets/Mitzi4132/VideoLLava
下载链接
链接失效反馈官方服务:
资源简介:
MVBench数据集引入了一种新颖的静态到动态方法,用于定义与时间相关的任务。通过将静态任务转换为动态任务,促进了需要从感知到认知的广泛时间能力的视频任务的系统生成。在任务定义的指导下,自动将公共视频注释转换为多项选择QA以进行任务评估。这种独特的范式使得MVBench的创建高效且人工干预最少,同时通过真实视频注释确保评估的公平性,并避免有偏见的LLM评分。数据集包含20个时间任务示例。
The MVBench dataset introduces a novel static-to-dynamic method for defining temporal-related tasks. By converting static tasks into dynamic ones, it systematically generates video tasks requiring a wide range of temporal abilities, from perception to cognition. The dataset evaluates tasks by automatically transforming public video annotations into multiple-choice QA, ensuring evaluation fairness. It includes 20 temporal task examples and provides an evaluation example.
提供机构:
Mitzi4132



