Huangzx1023/Video-Training

该数据集来自论文《Video-R1: Reinforcing Video Reasoning in MLLMs》，旨在增强多模态大语言模型（MLLMs）的视频推理能力。数据集包含视频和图像数据，其中视频数据文件夹包括CLEVRER、LLaVA-Video-178K、NeXT-QA、PerceptionTest和STAR，图像数据文件夹包括Chart、General、Knowledge、Math、OCR和Spatial。数据集提供了两个JSON文件：Video-R1-COT-165k.json用于监督微调（SFT）冷启动，Video-R1-260k.json用于强化学习（RL）训练。数据格式为JSON，每个样本包含问题ID、问题文本、数据类型（视频或图像）、问题类型（如多项选择）、选项列表、推理过程（chain-of-thought）、答案、文件路径和数据来源。示例展示了视频问答任务，涉及俄语屏幕内容的识别，并提供了详细的推理步骤和答案。This dataset is presented in the paper Video-R1: Reinforcing Video Reasoning in MLLMs and is designed to enhance video reasoning in multimodal large language models (MLLMs). It includes video and image data, with video data folders covering CLEVRER, LLaVA-Video-178K, NeXT-QA, PerceptionTest, and STAR, and image data folders covering Chart, General, Knowledge, Math, OCR, and Spatial. The dataset provides two JSON files: Video-R1-COT-165k.json for supervised fine-tuning (SFT) cold start, and Video-R1-260k.json for reinforcement learning (RL) training. The data format is JSON, with each sample containing a problem ID, problem text, data type (video or image), problem type (e.g., multiple choice), list of options, reasoning process (chain-of-thought), solution, file path, and data source. An example demonstrates a video question-answering task involving the identification of Russian on-screen content, with detailed reasoning steps and the answer provided.