The InternVideo3 long-video supervised fine-tuning data is available on Hugging Face:
Each sample contains a YouTube video id and QA annotations for detailed long-video description and reasoning. If you use this dataset, please cite:
@misc{yan2026internvideo3agentifyfoundationmodels,
title={InternVideo3: Agentify Foundation Models with Multimodal Contextual Reasoning}… See the full description on the dataset page:
https://huggingface.co/datasets/yanziang/InternVideo3_Dataset.