CoF-T2I: Video Models as Pure Visual Reasoners for Text-to-Image Generation

Chengzhuo Tong, Mingkun Chang, Shenglong Zhang, Yuran Wang, Cheng Liang, Zhizheng Zhao, Bohan Zeng, Yang Shi, Ruichuan An, Yifan Dai, Ziming Zhao, Guanbin Li, Pengfei Wan, Yuanxing Zhang, Wentao Zhang
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:121483-121498, 2026.

Abstract

Recent video generation models have revealed the emergence of Chain-of-Frame (CoF) reasoning, enabling frame-by-frame visual inference. With this capability, video models have been successfully applied to various visual tasks (e.g., maze solving, visual puzzles). However, their potential to enhance text-to-image (T2I) generation remains largely unexplored due to the absence of a clearly defined visual reasoning starting point and interpretable intermediate states in the T2I generation process. To bridge this gap, we propose CoF-T2I, a model that integrates CoF reasoning into T2I generation via progressive visual refinement, where intermediate frames act as explicit reasoning steps and the final frame is taken as output. To establish such explicit generation process, we curate CoF-Evol-Instruct, a dataset of CoF trajectories that model the generation process from semantics to aesthetics. To further improve quality and avoid motion artifacts, we enable an independent encoding operation for each frame. Experiments show that CoF-T2I significantly outperforms the base video model and achieves competitive performance, reaching 0.86 on GenEval and 7.468 on Imagine-Bench. These results indicate the substantial promise of video models for advancing high-quality text-to-image generation.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-tong26a, title = {{C}o{F}-{T}2{I}: Video Models as Pure Visual Reasoners for Text-to-Image Generation}, author = {Tong, Chengzhuo and Chang, Mingkun and Zhang, Shenglong and Wang, Yuran and Liang, Cheng and Zhao, Zhizheng and Zeng, Bohan and Shi, Yang and An, Ruichuan and Dai, Yifan and Zhao, Ziming and Li, Guanbin and Wan, Pengfei and Zhang, Yuanxing and Zhang, Wentao}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {121483--121498}, 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/tong26a/tong26a.pdf}, url = {https://proceedings.mlr.press/v306/tong26a.html}, abstract = {Recent video generation models have revealed the emergence of Chain-of-Frame (CoF) reasoning, enabling frame-by-frame visual inference. With this capability, video models have been successfully applied to various visual tasks (e.g., maze solving, visual puzzles). However, their potential to enhance text-to-image (T2I) generation remains largely unexplored due to the absence of a clearly defined visual reasoning starting point and interpretable intermediate states in the T2I generation process. To bridge this gap, we propose CoF-T2I, a model that integrates CoF reasoning into T2I generation via progressive visual refinement, where intermediate frames act as explicit reasoning steps and the final frame is taken as output. To establish such explicit generation process, we curate CoF-Evol-Instruct, a dataset of CoF trajectories that model the generation process from semantics to aesthetics. To further improve quality and avoid motion artifacts, we enable an independent encoding operation for each frame. Experiments show that CoF-T2I significantly outperforms the base video model and achieves competitive performance, reaching 0.86 on GenEval and 7.468 on Imagine-Bench. These results indicate the substantial promise of video models for advancing high-quality text-to-image generation.} }
Endnote
%0 Conference Paper %T CoF-T2I: Video Models as Pure Visual Reasoners for Text-to-Image Generation %A Chengzhuo Tong %A Mingkun Chang %A Shenglong Zhang %A Yuran Wang %A Cheng Liang %A Zhizheng Zhao %A Bohan Zeng %A Yang Shi %A Ruichuan An %A Yifan Dai %A Ziming Zhao %A Guanbin Li %A Pengfei Wan %A Yuanxing Zhang %A Wentao Zhang %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-tong26a %I PMLR %P 121483--121498 %U https://proceedings.mlr.press/v306/tong26a.html %V 306 %X Recent video generation models have revealed the emergence of Chain-of-Frame (CoF) reasoning, enabling frame-by-frame visual inference. With this capability, video models have been successfully applied to various visual tasks (e.g., maze solving, visual puzzles). However, their potential to enhance text-to-image (T2I) generation remains largely unexplored due to the absence of a clearly defined visual reasoning starting point and interpretable intermediate states in the T2I generation process. To bridge this gap, we propose CoF-T2I, a model that integrates CoF reasoning into T2I generation via progressive visual refinement, where intermediate frames act as explicit reasoning steps and the final frame is taken as output. To establish such explicit generation process, we curate CoF-Evol-Instruct, a dataset of CoF trajectories that model the generation process from semantics to aesthetics. To further improve quality and avoid motion artifacts, we enable an independent encoding operation for each frame. Experiments show that CoF-T2I significantly outperforms the base video model and achieves competitive performance, reaching 0.86 on GenEval and 7.468 on Imagine-Bench. These results indicate the substantial promise of video models for advancing high-quality text-to-image generation.
APA
Tong, C., Chang, M., Zhang, S., Wang, Y., Liang, C., Zhao, Z., Zeng, B., Shi, Y., An, R., Dai, Y., Zhao, Z., Li, G., Wan, P., Zhang, Y. & Zhang, W.. (2026). CoF-T2I: Video Models as Pure Visual Reasoners for Text-to-Image Generation. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:121483-121498 Available from https://proceedings.mlr.press/v306/tong26a.html.

Related Material