This organization is a synthetic data spring for video reasoning. It hosts the parameterized generators behind VBVR (Very Big Video Reasoning) — each one capable of producing unlimited training data on demand.
Every repository is a self-contained task generator: give it a seed and parameters, and it deterministically produces (first frame, prompt, final frame, ground-truth video) tuples. There is no fixed dataset size — every generator can mint fresh, unique samples indefinitely. Need 10K examples? 10M? Just run the generator longer.
For all non-generator repositories in this project, checkout https://github.com/Video-Reason
| Perception Sort Objects By Rule ![]() |
Transformation Combined Objects Spinning ![]() |
Spatiality Grid Shortest Path ![]() |
Abstraction Symmetry Random ![]() |
Knowledge Hit Target After Bounce ![]() |
![]() Maze |
![]() Mirror Reflection |
![]() Raven’s Matrices |
![]() Sliding Puzzle |
![]() Clock |
Every generator follows a standard interface:
Generator(seed, params) -> first_frame.png, prompt.txt, final_frame.png, ground_truth.mp4
Start building your own with template-data-generator or convert existing datasets with template-data-pipeline.
150 generators are open-source today, organized across five cognitive categories — with more on the way.
Extraction of structured representations from sensory input.
View all 45 generators
Distillation of generalizable knowledge from particular experiences.
View all 33 generators
Manipulation and synthesis of mental representations.
View all 29 generators
Propositional truths, either learned or innate.
View all 23 generators
Representation of places and geometric relationships.
View all 20 generators
- Fork
template-data-generator - Design your task targeting one of the five cognitive faculties
- Implement a parameterized generator that produces
first_frame.png,prompt.txt,final_frame.png, andground_truth.mp4 - Submit for review against six quality criteria: information sufficiency, deterministic solvability, video dependency, visual clarity, parametric diversity (10,000+ instances), and technical feasibility
- Website: Video-Reason.com
- Paper: A Very Big Video Reasoning Suite
- Slack: Join our workspace
- HuggingFace: Video-Reason
- Contact: hokinxqdeng@gmail.com
If you use VBVR in your research, please cite:
@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}
}









