Paper which allows implementing meanflow without changing the...

@SwayStar123
sway@SwayStar123
31 views Oct 30, 2025 ~1 min read
Advertisement
1
Paper which allows implementing meanflow without changing the architecture of a flow model, meaning you can finetune an existing flow model into being a meanflow model without arch changes now.

DECOUPLED MEANFLOW: TURNING FLOW MODELS INTO FLOW MAPS FOR ACCELERATED SAMPLING

Denoising generative models, such as diffusion and flow-based models, produce high-quality samples but require many denoising steps due to discretization error. Flow maps, which estimate the average velocity between timesteps, mitigate this error and enable faster sampling. However, their training typically demands architectural changes that limit compatibility with pretrained flow models. We introduce Decoupled MeanFlow, a simple decoding strategy that converts flow models into flow map models without architectural modifications. Our method conditions the final blocks of diffusion transformers on the subsequent timestep, allowing pretrained flow models to be directly repurposed as flow maps. Combined with enhanced training techniques, this design enables high-quality generation in as few as 1–4 steps. Notably, we find that training flow models and subsequently converting them is more efficient and effective than training flow maps from scratch. On ImageNet 256×256 and 512×512, our models attain 1-step FID of 2.16 and 2.12, respectively, surpassing prior art by a large margin. Furthermore, we achieve FID of 1.51 and 1.68 when increasing the steps to 4, which nearly matches the performance of flow models while delivering over 100× faster inference.

One of the authors (Sihyun Yu) is the main author of REPA
Media image
Actions
What You Can Do
  • Export as PDF or Markdown
  • Batch Export to Notion
  • Bookmark & Highlight
  • Screenshot Tweet
Create Free Account

Includes 7-day Premium trial

Advertisement