Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2512.11831 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912865644969984 |
|---|---|
| author | Lin, Haitao Hu, Peiyan Ren, Minsi Gao, Zhifeng Ma, Zhi-Ming ke, Guolin Wu, Tailin Li, Stan Z. |
| author_facet | Lin, Haitao Hu, Peiyan Ren, Minsi Gao, Zhifeng Ma, Zhi-Ming ke, Guolin Wu, Tailin Li, Stan Z. |
| contents | Recent advances in few-step diffusion models have demonstrated their efficiency and effectiveness by shortcutting the probabilistic paths of diffusion models, especially in training one-step diffusion models from scratch (\emph{a.k.a.} shortcut models). However, their theoretical derivation and practical implementation are often closely coupled, which obscures the design space. To address this, we propose a common design framework for representative shortcut models. This framework provides theoretical justification for their validity and disentangles concrete component-level choices, thereby enabling systematic identification of improvements. With our proposed improvements, the resulting one-step model achieves a new state-of-the-art FID50k of 2.85 on ImageNet-256x256 under the classifier-free guidance setting with one step generation, and further reaches FID50k of 2.53 with 2x training steps. Remarkably, the model requires no pre-training, distillation, or curriculum learning. We believe our work lowers the barrier to component-level innovation in shortcut models and facilitates principled exploration of their design space. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_11831 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | On the Design of One-step Diffusion via Shortcutting Flow Paths Lin, Haitao Hu, Peiyan Ren, Minsi Gao, Zhifeng Ma, Zhi-Ming ke, Guolin Wu, Tailin Li, Stan Z. Machine Learning Computer Vision and Pattern Recognition Recent advances in few-step diffusion models have demonstrated their efficiency and effectiveness by shortcutting the probabilistic paths of diffusion models, especially in training one-step diffusion models from scratch (\emph{a.k.a.} shortcut models). However, their theoretical derivation and practical implementation are often closely coupled, which obscures the design space. To address this, we propose a common design framework for representative shortcut models. This framework provides theoretical justification for their validity and disentangles concrete component-level choices, thereby enabling systematic identification of improvements. With our proposed improvements, the resulting one-step model achieves a new state-of-the-art FID50k of 2.85 on ImageNet-256x256 under the classifier-free guidance setting with one step generation, and further reaches FID50k of 2.53 with 2x training steps. Remarkably, the model requires no pre-training, distillation, or curriculum learning. We believe our work lowers the barrier to component-level innovation in shortcut models and facilitates principled exploration of their design space. |
| title | On the Design of One-step Diffusion via Shortcutting Flow Paths |
| topic | Machine Learning Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2512.11831 |