Improved Mean Flows: On the Challenges of Fastforward Generative Models
Zhengyang Geng, Yiyang Lu, Zongze Wu, Eli Shechtman, J. Zico Kolter, Kaiming He
Abstract
MeanFlow (MF) has recently been established as a framework for one-step generative modeling. However, its ``fastforward''nature introduces key challenges in both the training objective and the guidance mechanism. First, the original MF's training target depends not only on the underlying ground-truth fields but also on the network itself. To address this issue, we recast the objective as a loss on the instantaneous velocity , re-parameterized by a network that predicts the average velocity . Our reformulation yields a more standard regression problem and improves the training stability. Second, the original MF fixes the classifier-free guidance scale during training, which sacrifices flexibility. We tackle this issue by formulating guidance as explicit conditioning variables, thereby retaining flexibility at test time. The diverse conditions are processed through in-context conditioning, which reduces model size and benefits performance. Overall, our () method, trained entirely from scratch, achieves FID with a single function evaluation (1-NFE) on ImageNet 256256. iMF substantially outperforms prior methods of this kind and closes the gap with multi-step methods while using no distillation. We hope our work will further advance fastforward generative modeling as a stand-alone paradigm.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
15 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- High-Resolution Image Synthesis with Latent Diffusion Models cites Diffusion Models Beat GANs on Image Synthesis
- High-Resolution Image Synthesis with Latent Diffusion Models cites Denoising Diffusion Probabilistic Models
- Bi-Anchor Interpolation Solver for Accelerating Generative Modeling cites Scalable Diffusion Models with Transformers
- Bi-Anchor Interpolation Solver for Accelerating Generative Modeling cites Diffusion Models Beat GANs on Image Synthesis
- Bi-Anchor Interpolation Solver for Accelerating Generative Modeling cites Denoising Diffusion Probabilistic Models
- Scalable Diffusion Models with Transformers cites High-Resolution Image Synthesis with Latent Diffusion Models
- Scalable Diffusion Models with Transformers cites Elucidating the Design Space of Diffusion-Based Generative Models
- Scalable Diffusion Models with Transformers cites Diffusion Models Beat GANs on Image Synthesis
- Scalable Diffusion Models with Transformers cites Denoising Diffusion Probabilistic Models
- Elucidating the Design Space of Diffusion-Based Generative Models cites High-Resolution Image Synthesis with Latent Diffusion Models
- Elucidating the Design Space of Diffusion-Based Generative Models cites Diffusion Models Beat GANs on Image Synthesis
- Elucidating the Design Space of Diffusion-Based Generative Models cites Denoising Diffusion Probabilistic Models
- Diffusion Models Beat GANs on Image Synthesis cites Denoising Diffusion Probabilistic Models
- Bidirectional Normalizing Flow: From Data to Noise and Back cites High-Resolution Image Synthesis with Latent Diffusion Models
- Bidirectional Normalizing Flow: From Data to Noise and Back cites Scalable Diffusion Models with Transformers
- Bidirectional Normalizing Flow: From Data to Noise and Back cites Diffusion Models Beat GANs on Image Synthesis
- Bidirectional Normalizing Flow: From Data to Noise and Back cites Denoising Diffusion Probabilistic Models
- Categorical Flow Maps cites Scalable Diffusion Models with Transformers
- Learning Hamiltonian Flow Maps: Mean Flow Consistency for Large-Timestep Molecular Dynamics cites Denoising Diffusion Probabilistic Models
Cited by top-tier papers15
- Categorical Flow MapsDaan Roos, Oscar Davis, Floor Eijkelboom, Michael Bronstein et al.ICML 2026 · 23 citations
- Bidirectional Normalizing Flow: From Data to Noise and BackYiyang Lu, Qiao Sun, Xianbang Wang, Zhicheng Jiang et al.CVPR 2026 · 7 citations
- Bi-Anchor Interpolation Solver for Accelerating Generative ModelingHongxu CHEN, Hongxiang Li, Zhen Wang, Long ChenICML 2026 · 3 citations
- WFR-MFM: One-Step Inference for Dynamic Unbalanced Optimal TransportXinyu Wang, Ruoyu Wang, Qiangwei Peng, Peijie Zhou et al.ICML 2026 · 3 citations
- Learning Hamiltonian Flow Maps: Mean Flow Consistency for Large-Timestep Molecular DynamicsWinfried Ripken, Michael Plainer, Gregor Lied, Thorben Frank et al.ICML 2026 · 3 citations
Builds on40
- Denoising Diffusion Probabilistic ModelsJonathan Ho, Ajay Jain, Pieter AbbeelNeurIPS 2020 · 35,902 citations
- Diffusion Models Beat GANs on Image SynthesisPrafulla Dhariwal, Alexander Quinn NicholNeurIPS 2021 · 13,211 citations
- High-Resolution Image Synthesis with Latent Diffusion ModelsRobin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser et al.CVPR 2022 · 13,123 citations
- Scalable Diffusion Models with TransformersWilliam Peebles, Saining XieICCV 2023 · 5,568 citations
- Elucidating the Design Space of Diffusion-Based Generative ModelsTero Karras, Miika Aittala, Timo Aila, Samuli LaineNeurIPS 2022 · 3,959 citations
Related papers
- Mean Flows for One-step Generative ModelingZhengyang Geng, Mingyang Deng, Xingjian Bai, Zico Kolter et al.NeurIPS 2025 · 628 citations
- Understanding, Accelerating, and Improving MeanFlow TrainingJin-Young Kim, Hyojun Go, Lea Bogensperger, Julius Erbach et al.CVPR 2026 · 4 citations
- SoFlow: Solution Flow Models for One-Step Generative ModelingTianze Luo, Haotian Yuan, Zhuang LiuICLR 2026 · 18 citations
- One-step Latent-free Image Generation with Pixel Mean FlowsYiyang Lu, Susie Lu, Qiao Sun, Hanhong Zhao et al.ICML 2026
- MeanFlow Transformers with Representation AutoencodersZheyuan Hu, Chieh-Hsin Lai, Ge Wu, Yuki Mitsufuji et al.CVPR 2026 · 6 citations
