A Very Big Video Reasoning Suite

Maijunxian Wang, Ruisi Wang, Juyi Lin, Ran Ji, Thaddäus Wiedemer, Qingying Gao, Dezhi Luo, Yaoyao Qian, Lianyu Huang, Zelong Hong, Jiahui Ge, Qianli Ma, Hang He, Yifan Zhou, Lingzi Guo, Lantao Mei, Jiachen Li, Hanwen Xing, Tianqi Zhao, Yu Fengyuan, Weihang Xiao, Yizheng Jiao, Jianheng Hou, Danyang Zhang, Pengcheng Xu, Boyang Zhong, Zehong Zhao, Gaoyun Fang, John Kitaoka, Xu Yile, Hua Xu, Kenton Blacutt, Tin Nguyen, Siyuan Song, Haoran Sun, Shaoyue Wen, Linyang He, Runming Wang, Yanzhi Wang, Mengyue Yang, Ziqiao Ma, Raphaël Millière, Freda Shi, Nuno Vasconcelos, Daniel Khashabi, Alan Yuille, Yilun Du, Ziming Liu, Dahua Lin, Ziwei Liu, Vikash Kumar, Yijiang Li, Lei Yang, Zhongang Cai, Hokin Deng
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:130550-130588, 2026.

Abstract

Rapid progress in video models has largely focused on visual quality, leaving their reasoning capabilities underexplored. Video reasoning grounds intelligence in spatiotemporally consistent visual environments that go beyond what text can naturally capture, enabling intuitive reasoning over spatiotemporal structure, such as continuity, interaction, and causality. However, systematically studying video reasoning and its scaling behavior is hindered by the lack of large-scale video reasoning training data. To address this gap, we introduce the Very Big Video Reasoning (VBVR) Dataset, an unprecedentedly large-scale resource spanning 200 curated reasoning tasks following a principled taxonomy, and over one million video clips, making it approximately three orders of magnitude larger than existing datasets. We further present VBVR-Bench, a verifiable evaluation framework that moves beyond model-based judging by incorporating rule-based, human-aligned scorers, enabling reproducible and interpretable diagnosis of video reasoning capabilities. Leveraging the VBVR suite, we conduct one of the first video reasoning scaling studies and observe early signs of emergent generalization to unseen reasoning tasks. Together, VBVR lays a foundation for the next stage of research in generalizable video reasoning. The data, benchmark tool kit, and models are released publicly at video-reason.com.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-wang26ia, 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 Fengyuan, Yu 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 Yile, Xu 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}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {130550--130588}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/wang26ia/wang26ia.pdf}, url = {https://proceedings.mlr.press/v306/wang26ia.html}, abstract = {Rapid progress in video models has largely focused on visual quality, leaving their reasoning capabilities underexplored. Video reasoning grounds intelligence in spatiotemporally consistent visual environments that go beyond what text can naturally capture, enabling intuitive reasoning over spatiotemporal structure, such as continuity, interaction, and causality. However, systematically studying video reasoning and its scaling behavior is hindered by the lack of large-scale video reasoning training data. To address this gap, we introduce the Very Big Video Reasoning (VBVR) Dataset, an unprecedentedly large-scale resource spanning 200 curated reasoning tasks following a principled taxonomy, and over one million video clips, making it approximately three orders of magnitude larger than existing datasets. We further present VBVR-Bench, a verifiable evaluation framework that moves beyond model-based judging by incorporating rule-based, human-aligned scorers, enabling reproducible and interpretable diagnosis of video reasoning capabilities. Leveraging the VBVR suite, we conduct one of the first video reasoning scaling studies and observe early signs of emergent generalization to unseen reasoning tasks. Together, VBVR lays a foundation for the next stage of research in generalizable video reasoning. The data, benchmark tool kit, and models are released publicly at video-reason.com.} }
Endnote
%0 Conference Paper %T A Very Big Video Reasoning Suite %A Maijunxian Wang %A Ruisi Wang %A Juyi Lin %A Ran Ji %A Thaddäus Wiedemer %A Qingying Gao %A Dezhi Luo %A Yaoyao Qian %A Lianyu Huang %A Zelong Hong %A Jiahui Ge %A Qianli Ma %A Hang He %A Yifan Zhou %A Lingzi Guo %A Lantao Mei %A Jiachen Li %A Hanwen Xing %A Tianqi Zhao %A Yu Fengyuan %A Weihang Xiao %A Yizheng Jiao %A Jianheng Hou %A Danyang Zhang %A Pengcheng Xu %A Boyang Zhong %A Zehong Zhao %A Gaoyun Fang %A John Kitaoka %A Xu Yile %A Hua Xu %A Kenton Blacutt %A Tin Nguyen %A Siyuan Song %A Haoran Sun %A Shaoyue Wen %A Linyang He %A Runming Wang %A Yanzhi Wang %A Mengyue Yang %A Ziqiao Ma %A Raphaël Millière %A Freda Shi %A Nuno Vasconcelos %A Daniel Khashabi %A Alan Yuille %A Yilun Du %A Ziming Liu %A Dahua Lin %A Ziwei Liu %A Vikash Kumar %A Yijiang Li %A Lei Yang %A Zhongang Cai %A Hokin Deng %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-wang26ia %I PMLR %P 130550--130588 %U https://proceedings.mlr.press/v306/wang26ia.html %V 306 %X Rapid progress in video models has largely focused on visual quality, leaving their reasoning capabilities underexplored. Video reasoning grounds intelligence in spatiotemporally consistent visual environments that go beyond what text can naturally capture, enabling intuitive reasoning over spatiotemporal structure, such as continuity, interaction, and causality. However, systematically studying video reasoning and its scaling behavior is hindered by the lack of large-scale video reasoning training data. To address this gap, we introduce the Very Big Video Reasoning (VBVR) Dataset, an unprecedentedly large-scale resource spanning 200 curated reasoning tasks following a principled taxonomy, and over one million video clips, making it approximately three orders of magnitude larger than existing datasets. We further present VBVR-Bench, a verifiable evaluation framework that moves beyond model-based judging by incorporating rule-based, human-aligned scorers, enabling reproducible and interpretable diagnosis of video reasoning capabilities. Leveraging the VBVR suite, we conduct one of the first video reasoning scaling studies and observe early signs of emergent generalization to unseen reasoning tasks. Together, VBVR lays a foundation for the next stage of research in generalizable video reasoning. The data, benchmark tool kit, and models are released publicly at video-reason.com.
APA
Wang, M., Wang, R., Lin, J., Ji, R., Wiedemer, T., Gao, Q., Luo, D., Qian, Y., Huang, L., Hong, Z., Ge, J., Ma, Q., He, H., Zhou, Y., Guo, L., Mei, L., Li, J., Xing, H., Zhao, T., Fengyuan, Y., Xiao, W., Jiao, Y., Hou, J., Zhang, D., Xu, P., Zhong, B., Zhao, Z., Fang, G., Kitaoka, J., Yile, X., Xu, H., Blacutt, K., Nguyen, T., Song, S., Sun, H., Wen, S., He, L., Wang, R., Wang, Y., Yang, M., Ma, Z., Millière, R., Shi, F., Vasconcelos, N., Khashabi, D., Yuille, A., Du, Y., Liu, Z., Lin, D., Liu, Z., Kumar, V., Li, Y., Yang, L., Cai, Z. & Deng, H.. (2026). A Very Big Video Reasoning Suite. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:130550-130588 Available from https://proceedings.mlr.press/v306/wang26ia.html.

Related Material