Runge-Kutta Approximation and Decoupled Attention for Rectified Flow Inversion and Semantic Editing

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Chen, Weiming, Zhu, Zhihan, Wang, Yijia, He, Zhihai
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866918142129733632
author Chen, Weiming
Zhu, Zhihan
Wang, Yijia
He, Zhihai
author_facet Chen, Weiming
Zhu, Zhihan
Wang, Yijia
He, Zhihai
contents Rectified flow (RF) models have recently demonstrated superior generative performance compared to DDIM-based diffusion models. However, in real-world applications, they suffer from two major challenges: (1) low inversion accuracy that hinders the consistency with the source image, and (2) entangled multimodal attention in diffusion transformers, which hinders precise attention control. To address the first challenge, we propose an efficient high-order inversion method for rectified flow models based on the Runge-Kutta solver of differential equations. To tackle the second challenge, we introduce Decoupled Diffusion Transformer Attention (DDTA), a novel mechanism that disentangles text and image attention inside the multimodal diffusion transformers, enabling more precise semantic control. Extensive experiments on image reconstruction and text-guided editing tasks demonstrate that our method achieves state-of-the-art performance in terms of fidelity and editability. Code is available at https://github.com/wmchen/RKSovler_DDTA.
format Preprint
id arxiv_https___arxiv_org_abs_2509_12888
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Runge-Kutta Approximation and Decoupled Attention for Rectified Flow Inversion and Semantic Editing
Chen, Weiming
Zhu, Zhihan
Wang, Yijia
He, Zhihai
Computer Vision and Pattern Recognition
Artificial Intelligence
Rectified flow (RF) models have recently demonstrated superior generative performance compared to DDIM-based diffusion models. However, in real-world applications, they suffer from two major challenges: (1) low inversion accuracy that hinders the consistency with the source image, and (2) entangled multimodal attention in diffusion transformers, which hinders precise attention control. To address the first challenge, we propose an efficient high-order inversion method for rectified flow models based on the Runge-Kutta solver of differential equations. To tackle the second challenge, we introduce Decoupled Diffusion Transformer Attention (DDTA), a novel mechanism that disentangles text and image attention inside the multimodal diffusion transformers, enabling more precise semantic control. Extensive experiments on image reconstruction and text-guided editing tasks demonstrate that our method achieves state-of-the-art performance in terms of fidelity and editability. Code is available at https://github.com/wmchen/RKSovler_DDTA.
title Runge-Kutta Approximation and Decoupled Attention for Rectified Flow Inversion and Semantic Editing
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2509.12888