OpenMOSS-Team/Realtime-QA-100K
收藏资源简介:
Realtime-QA-100K是一个包含10万个样本的实时视频问答数据集,基于YouTube视频构建。每个样本包含一个多模态对话和帧时间戳元数据,这些元数据将助手文本中的每个`<|video|>`令牌与一个视频帧时间戳对齐。该数据集是MOSS-Video-Preview实时训练数据的开源子集,旨在支持开源社区中实时推理模型的训练。数据集中包含93,815个英语样本和6,185个中文样本,覆盖27个类别,视频时长在4分钟以内,帧采样率为1 fps。数据集仅提供注释、YouTube视频ID和时间戳元数据,不包含实际视频文件。
Realtime-QA-100K is a 100K-sample realtime video question answering dataset constructed from YouTube videos. Each sample contains a multimodal conversation and frame timestamp metadata that aligns every `<|video|>` token in the assistant text with one video frame timestamp. It is the open-source subset of the real-time training data for MOSS-Video-Preview, released to support training real-time reasoning models in the open-source community. The dataset includes 93,815 English samples and 6,185 Chinese samples, covering 27 categories, with video durations within 4 minutes and frame sampling at 1 fps. It provides annotations, YouTube video IDs, and timestamp metadata, but does not include actual video files.




