![[2602.11850] Free Lunch for Stabilizing Rectified Flow Inversion](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.11850] Free Lunch for Stabilizing Rectified Flow Inversion
Summary
This paper presents Proximal-Mean Inversion (PMI), a novel method for stabilizing Rectified-Flow (RF) models, enhancing image reconstruction and editing quality while reducing errors in generative tasks.
Why It Matters
The advancement of generative models like RF has significant implications for fields such as computer vision and machine learning. This research addresses critical stability issues in RF-based inversion methods, which can improve the reliability and effectiveness of applications in image generation and editing. By proposing a training-free approach, the authors contribute to the ongoing evolution of generative modeling techniques, making them more accessible and efficient.
Key Takeaways
- Proximal-Mean Inversion (PMI) stabilizes velocity fields in RF models.
- The method improves image reconstruction and editing quality.
- PMI reduces approximation errors that affect generative tasks.
- Mimic-CFG offers a lightweight solution for editing tasks.
- The proposed techniques achieve state-of-the-art performance on PIE-Bench.
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.11850 (cs) [Submitted on 12 Feb 2026 (v1), last revised 13 Feb 2026 (this version, v2)] Title:Free Lunch for Stabilizing Rectified Flow Inversion Authors:Chenru Wang, Beier Zhu, Chi Zhang View a PDF of the paper titled Free Lunch for Stabilizing Rectified Flow Inversion, by Chenru Wang and 2 other authors View PDF HTML (experimental) Abstract:Rectified-Flow (RF)-based generative models have recently emerged as strong alternatives to traditional diffusion models, demonstrating state-of-the-art performance across various tasks. By learning a continuous velocity field that transforms simple noise into complex data, RF-based models not only enable high-quality generation, but also support training-free inversion, which facilitates downstream tasks such as reconstruction and editing. However, existing inversion methods, such as vanilla RF-based inversion, suffer from approximation errors that accumulate across timesteps, leading to unstable velocity fields and degraded reconstruction and editing quality. To address this challenge, we propose Proximal-Mean Inversion (PMI), a training-free gradient correction method that stabilizes the velocity field by guiding it toward a running average of past velocities, constrained within a theoretically derived spherical Gaussian. Furthermore, we introduce mimic-CFG, a lightweight velocity correction scheme for editing tasks, which interpolates between the current velocit...
