Multi-Resolution Flow Matching: Training-Free Diffusion Acceleration via Staged Sampling
Quick Answer
MrFlow introduces a training-free multi-resolution acceleration strategy for flow-matching models, achieving 10x inference speedup while maintaining quality within 1% of prior performance.
Quick Take
This method utilizes a staged pipeline that combines low-resolution structure generation with high-resolution refinement, outperforming existing training-free strategies and allowing for further acceleration when combined with timestep distillation techniques.
Key Points
- Achieves 10x end-to-end acceleration without any training or runtime identification.
- Maintains OneIG performance within a 1% gap compared to pre-acceleration.
- Combines low-resolution sampling with high-frequency resampling for detail refinement.
- Can be orthogonally combined with timestep distillation for up to 25x acceleration.
- Utilizes a lightweight GAN-based model for super-resolution in pixel space.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Hardware-agnostic strategies for accelerating text-to-image diffusion, such as timestep distillation and feature caching, can reduce inference time without custom kernels or system-level optimization. Among them, multi-resolution generation strategies have recently received broad attention, attaining more than 5x speedup without any training. However, the design of performing upsampling in the latent space, together with the selective modification of partial regions, causes these methods to exhibit noticeable blurring or artifacts. To this end, we propose MrFlow, a training-free multi-resolution acceleration strategy for pretrained flow-matching models built upon a staged low-to-high-resolution pipeline. MrFlow first rapidly generates the main structure at low resolution, then performs super-resolution in the pixel space using a lightweight pretrained GAN-based model, subsequently injects low-strength noise to enable high-frequency resampling, and finally refines the details at high resolution. Quantitative and qualitative results on FLUX.1-dev and Qwen-Image show that MrFlow exploits the quadratic token reduction and reduced step requirement of low-resolution sampling to achieve 10x end-to-end acceleration while keeping OneIG within a 1% gap relative to that before acceleration, significantly surpassing other training-free acceleration strategies, and requiring no training or runtime dynamic identification whatsoever. MrFlow can further be directly combined orthogonally with pre-trained timestep distillation strategies, achieving even higher generation acceleration of up to 25x.
| Comments: | The code is available at this https URL |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2607.01642 [cs.CV] |
| (or arXiv:2607.01642v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2607.01642 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Xingyu Zheng [view email]
[v1]
Thu, 2 Jul 2026 03:14:57 UTC (7,892 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →ProMoE-FL: Prototype-conditioned Mixture of Experts for Multimodal Federated Learning with Missing Modalities
ProMoE-FL introduces a Prototype-conditioned Mixture-of-Experts framework for multimodal federated learning, effectively addressing missing modalities. It outperforms existing methods on four chest X-ray datasets, demonstrating superior feature synthesis capabilities in both homogeneous and heterogeneous settings.