Spaces:
Running
Running
| title: Video-Reason | |
| emoji: "\U0001F3AC" | |
| colorFrom: blue | |
| colorTo: purple | |
| sdk: static | |
| pinned: false | |
| tags: | |
| - video-reasoning | |
| - benchmark | |
| - video-generation | |
| - reasoning | |
| <div align="center"> | |
| ### We bet on a future where video reasoning is the next fundamental intelligence paradigm. | |
| [🌐 Website](https://video-reason.com) · [💻 GitHub](https://github.com/VBVR-DataFactory) | |
| </div> | |
| ## About Us | |
| **Video-Reason** is a research initiative dedicated to advancing **video reasoning** as the next foundational intelligence paradigm — where spatiotemporal, embodied world experiences are more naturally captured than through text alone. We build large-scale datasets, benchmarks, and models to systematically study and scale video reasoning capabilities. | |
| ## VBVR: A Very Big Video Reasoning Suite | |
| Our flagship project, **VBVR (Very Big Video Reasoning)**, introduces an unprecedentedly large-scale resource for video reasoning research: | |
| - **200** curated reasoning tasks across 5 domains: **Perception**, **Abstraction**, **Spatiality**, **Transformation**, and **Knowledge** | |
| - **1,000,000+** video clips — approximately three orders of magnitude larger than existing datasets | |
| - **Verifiable evaluation** via rule-based, human-aligned scorers (no model-based judging) | |
| - **Early signs of emergent generalization** to unseen reasoning tasks through large-scale scaling studies | |
| ## Releases | |
| | Resource | Description | Link | | |
| |---|---|---| | |
| | VBVR-Wan2.2 | Strong baseline model fine-tuned from Wan2.2-I2V-A14B on the VBVR Dataset | [Model](https://huggingface.co/Video-Reason/VBVR-Wan2.2) | | |
| | VBVR-Dataset | 1M video reasoning training samples across 100 curated task generators (~370 GB) | [Dataset](https://huggingface.co/datasets/Video-Reason/VBVR-Dataset) | | |
| | VBVR-Bench-Data | Official benchmark test set (500 samples across in-domain and out-of-domain splits) | [Dataset](https://huggingface.co/datasets/Video-Reason/VBVR-Bench-Data) | | |
| | VBVR-Bench-Leaderboard | Public leaderboard for standardized model comparison | [Space](https://huggingface.co/spaces/Video-Reason/VBVR-Bench-Leaderboard) | | |
| ## Links | |
| - **Website**: https://video-reason.com | |
| - **Paper**: https://arxiv.org/abs/2602.20159 | |
| - **Slack**: https://join.slack.com/t/video-reason/shared_invite/zt-3qqf23icm-UC29fatWWYsIuzRNBR1lgg | |
| - **Hugging Face**: https://huggingface.co/Video-Reason | |
| - **Contact**: hokinxqdeng@gmail.com | |
| ## Citation | |
| If you use VBVR in your research, please cite: | |
| ```bibtex | |
| @article{vbvr2026, | |
| title = {A Very Big Video Reasoning Suite}, | |
| author = {Wang, Maijunxian and Wang, Ruisi and Lin, Juyi and Ji, Ran and | |
| Wiedemer, Thadd{\"a}us and Gao, Qingying and Luo, Dezhi and | |
| Qian, Yaoyao and Huang, Lianyu and Hong, Zelong and Ge, Jiahui and | |
| Ma, Qianli and He, Hang and Zhou, Yifan and Guo, Lingzi and | |
| Mei, Lantao and Li, Jiachen and Xing, Hanwen and Zhao, Tianqi and | |
| Yu, Fengyuan and Xiao, Weihang and Jiao, Yizheng and | |
| Hou, Jianheng and Zhang, Danyang and Xu, Pengcheng and | |
| Zhong, Boyang and Zhao, Zehong and Fang, Gaoyun and Kitaoka, John and | |
| Xu, Yile and Xu, Hua and Blacutt, Kenton and Nguyen, Tin and | |
| Song, Siyuan and Sun, Haoran and Wen, Shaoyue and He, Linyang and | |
| Wang, Runming and Wang, Yanzhi and Yang, Mengyue and Ma, Ziqiao and | |
| Milli{\`e}re, Rapha{\"e}l and Shi, Freda and Vasconcelos, Nuno and | |
| Khashabi, Daniel and Yuille, Alan and Du, Yilun and Liu, Ziming and | |
| Lin, Dahua and Liu, Ziwei and Kumar, Vikash and Li, Yijiang and | |
| Yang, Lei and Cai, Zhongang and Deng, Hokin}, | |
| journal = {arXiv preprint arXiv:2602.20159}, | |
| year = {2026}, | |
| url = {https://arxiv.org/abs/2602.20159} | |
| } | |
| ``` | |