Representation Entanglement for Generation: Training Diffusion Transformers Is Much Easier Than You Think
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866909811932659712 |
|---|---|
| author | Wu, Ge Zhang, Shen Shi, Ruijing Gao, Shanghua Chen, Zhenyuan Wang, Lei Chen, Zhaowei Gao, Hongcheng Tang, Yao Yang, Jian Cheng, Ming-Ming Li, Xiang |
| author_facet | Wu, Ge Zhang, Shen Shi, Ruijing Gao, Shanghua Chen, Zhenyuan Wang, Lei Chen, Zhaowei Gao, Hongcheng Tang, Yao Yang, Jian Cheng, Ming-Ming Li, Xiang |
| contents | REPA and its variants effectively mitigate training challenges in diffusion models by incorporating external visual representations from pretrained models, through alignment between the noisy hidden projections of denoising networks and foundational clean image representations. We argue that the external alignment, which is absent during the entire denoising inference process, falls short of fully harnessing the potential of discriminative representations. In this work, we propose a straightforward method called Representation Entanglement for Generation (REG), which entangles low-level image latents with a single high-level class token from pretrained foundation models for denoising. REG acquires the capability to produce coherent image-class pairs directly from pure noise, substantially improving both generation quality and training efficiency. This is accomplished with negligible additional inference overhead, requiring only one single additional token for denoising (<0.5\% increase in FLOPs and latency). The inference process concurrently reconstructs both image latents and their corresponding global semantics, where the acquired semantic knowledge actively guides and enhances the image generation process. On ImageNet 256$\times$256, SiT-XL/2 + REG demonstrates remarkable convergence acceleration, achieving $\textbf{63}\times$ and $\textbf{23}\times$ faster training than SiT-XL/2 and SiT-XL/2 + REPA, respectively. More impressively, SiT-L/2 + REG trained for merely 400K iterations outperforms SiT-XL/2 + REPA trained for 4M iterations ($\textbf{10}\times$ longer). Code is available at: https://github.com/Martinser/REG. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_01467 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Representation Entanglement for Generation: Training Diffusion Transformers Is Much Easier Than You Think Wu, Ge Zhang, Shen Shi, Ruijing Gao, Shanghua Chen, Zhenyuan Wang, Lei Chen, Zhaowei Gao, Hongcheng Tang, Yao Yang, Jian Cheng, Ming-Ming Li, Xiang Computer Vision and Pattern Recognition REPA and its variants effectively mitigate training challenges in diffusion models by incorporating external visual representations from pretrained models, through alignment between the noisy hidden projections of denoising networks and foundational clean image representations. We argue that the external alignment, which is absent during the entire denoising inference process, falls short of fully harnessing the potential of discriminative representations. In this work, we propose a straightforward method called Representation Entanglement for Generation (REG), which entangles low-level image latents with a single high-level class token from pretrained foundation models for denoising. REG acquires the capability to produce coherent image-class pairs directly from pure noise, substantially improving both generation quality and training efficiency. This is accomplished with negligible additional inference overhead, requiring only one single additional token for denoising (<0.5\% increase in FLOPs and latency). The inference process concurrently reconstructs both image latents and their corresponding global semantics, where the acquired semantic knowledge actively guides and enhances the image generation process. On ImageNet 256$\times$256, SiT-XL/2 + REG demonstrates remarkable convergence acceleration, achieving $\textbf{63}\times$ and $\textbf{23}\times$ faster training than SiT-XL/2 and SiT-XL/2 + REPA, respectively. More impressively, SiT-L/2 + REG trained for merely 400K iterations outperforms SiT-XL/2 + REPA trained for 4M iterations ($\textbf{10}\times$ longer). Code is available at: https://github.com/Martinser/REG. |
| title | Representation Entanglement for Generation: Training Diffusion Transformers Is Much Easier Than You Think |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2507.01467 |