Full search page →

Pixel MeanFlow — one-step pixels through a denoised field

Optional · paper guide 6 of 6 for ML 08 · Diffusion Models and Flow Matching. Citation numbers refer to the reference list at the bottom of this page.

Navigate: this page and this track: Navigate: this page and this track

Question and thesis

Can the interval learning of MeanFlow and the image-output design of JiT be combined into one-call pixel generation? Lu et al. introduce pixel MeanFlow, abbreviated pMF, by separating the network’s output space from its training loss space [50]. Read MeanFlow, iMF and JiT first. The connection is not obtained by asking an ordinary JiT checkpoint to use one Euler step; the two-time field and training objective change.

Method and figure

The original figure compares an average-velocity field with its induced denoised-image field along a simulated ODE trajectory. The image-like intermediate field is a modeling hypothesis, not a guarantee for every time pair.
Source: Lu et al., ICML 2026, PMLR 306 proceedings; official PMLR GitHub-hosted PDF. Figure 1, PDF p. 1. The original figure compares an average-velocity field with its induced denoised-image field along a simulated ODE trajectory. The image-like intermediate field is a modeling hypothesis, not a guarantee for every time pair. [50]

Use data at t=0t=0 and noise at t=1t=1, as in MeanFlow. For state ztz_t, define an image-like field and its network parameterization:

x⋆(zt,r,t)=zt−t u(zt,r,t),uθ(zt,r,t)=zt−xθ(zt,r,t)t.x^\star(z_t,r,t)=z_t-t\,u(z_t,r,t),\qquad u_\theta(z_t,r,t)=\frac{z_t-x_\theta(z_t,r,t)}{t}.

Here rr and tt are interval endpoints, uu is the underlying average-velocity field, x⋆x^\star is the induced denoised-image field, and xθx_\theta is the network’s direct prediction. The conversion uses t>0t\gt0 with implementation safeguards near zero. Feed this derived uθu_\theta into iMF’s compound instantaneous-velocity predictor and train against the conditional velocity ϵ−x\epsilon-x. Here xx is the sampled clean image and ϵ\epsilon is prior noise. Thus the route is image output → average velocity → instantaneous-velocity loss.

At r=tr=t, the ideal image-like field agrees with the denoising quantity associated with the instantaneous velocity. At r=0r=0, it equals the ODE endpoint. For intermediate pairs, the paper presents an approximate manifold hypothesis supported by Figure 1, not a theorem that all such outputs lie on the natural-image manifold. At the full interval, the sampling formula simplifies to a direct image prediction from noise.

Evidence

Table 2 compares image output with average-velocity output using fixed token sequence length. At ImageNet 256×256 with a B/16 model, 160 epochs, the specified optimizer and MSE objective, FID is 9.56 versus 164.89. At 64×64 with B/4, the two outputs instead give 3.80 and 3.82. These matched comparisons expose the high-dimensional prediction issue. Figure 3 separately studies optimization and perceptual-loss additions; those gains should not all be attributed to the field conversion.

Design analysis

The direct image output makes an image-space perceptual loss possible during generator training. The paper’s stronger recipe uses pretrained feature extractors for that loss, even though generation has no latent tokenizer or decoder. Therefore “latent-free” does not mean every component used in training is pretrained-free. Likewise, avoiding decoder latency and reducing network calls are distinct efficiency gains; actual runtime depends on model size and hardware.

Reading checkpoint

Substitute r=0r=0 and t=1t=1 into the MeanFlow update and the image-to-velocity conversion. Show why the final image equals the network output. Then explain why the interior field x⋆(zt,r,t)x^\star(z_t,r,t) is generally not identical to the clean training image paired with that noisy input.

Limits and next step

The evidence concerns the studied class-conditioned image setting, not arbitrary editing, text-to-image or video. The manifold argument is approximate, and a perceptual objective can trade exact fidelity for feature similarity. Return to the chapter’s evaluation section: compare full pipelines with matched resolution, conditioning, NFE, pretraining disclosures, quality and measured latency.

Continue reading: ML 08 · Evaluate generation

References

  1. L. Yang, Z. Zhang, Y. Song, S. Hong, R. Xu, Y. Zhao, W. Zhang, B. Cui and M.-H. Yang, “Diffusion Models: A Comprehensive Survey of Methods and Applications,” ACM Computing Surveys, 2023. arXiv
  2. Y. Lipman, M. Havasi, P. Holderrieth, N. Shaul, M. Le, B. Karrer, R. T. Q. Chen, D. Lopez-Paz, H. Ben-Hamu and I. Gat, “Flow Matching Guide and Code,” arXiv:2412.06264, 2024. arXiv
  3. J. Ho, A. Jain and P. Abbeel, “Denoising Diffusion Probabilistic Models,” NeurIPS, 2020. arXiv
  4. Y. Song, J. Sohl-Dickstein, D. P. Kingma, A. Kumar, S. Ermon and B. Poole, “Score-Based Generative Modeling through Stochastic Differential Equations,” ICLR, 2021. arXiv
  5. J. Song, C. Meng and S. Ermon, “Denoising Diffusion Implicit Models,” ICLR, 2021. arXiv
  6. J. Ho and T. Salimans, “Classifier-Free Diffusion Guidance,” NeurIPS 2021 Workshop on Deep Generative Models and Downstream Applications; arXiv:2207.12598, 2022. arXiv
  7. R. Rombach, A. Blattmann, D. Lorenz, P. Esser and B. Ommer, “High-Resolution Image Synthesis with Latent Diffusion Models,” CVPR, 2022. arXiv
  8. W. Peebles and S. Xie, “Scalable Diffusion Models with Transformers,” ICCV, 2023. arXiv
  9. P. Esser, S. Kulal, A. Blattmann et al., “Scaling Rectified Flow Transformers for High-Resolution Image Synthesis,” ICML, PMLR 235, 2024. PMLR
  10. Y. Lipman, R. T. Q. Chen, H. Ben-Hamu, M. Nickel and M. Le, “Flow Matching for Generative Modeling,” ICLR, 2023. arXiv
  11. X. Liu, C. Gong and Q. Liu, “Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow,” ICLR, 2023. arXiv
  12. Y. Song, P. Dhariwal, M. Chen and I. Sutskever, “Consistency Models,” ICML, 2023. arXiv
  13. Z. Geng, M. Deng, X. Bai, J. Z. Kolter and K. He, “Mean Flows for One-step Generative Modeling,” NeurIPS, 2025. arXiv
  14. S. J. D. Prince, Understanding Deep Learning, MIT Press, 2023. book site
  15. J. Sohl-Dickstein, E. Weiss, N. Maheswaranathan and S. Ganguli, “Deep Unsupervised Learning using Nonequilibrium Thermodynamics,” ICML, PMLR 37, 2015. PMLR
  16. P. Vincent, “A Connection Between Score Matching and Denoising Autoencoders,” Neural Computation, vol. 23, no. 7, pp. 1661–1674, 2011. doi
  17. Y. Song and S. Ermon, “Generative Modeling by Estimating Gradients of the Data Distribution,” NeurIPS, 2019. arXiv
  18. P. Dhariwal and A. Nichol, “Diffusion Models Beat GANs on Image Synthesis,” NeurIPS, 2021. NeurIPS
  19. A. Q. Nichol and P. Dhariwal, “Improved Denoising Diffusion Probabilistic Models,” ICML, PMLR 139, 2021. PMLR
  20. T. Karras, M. Aittala, T. Aila and S. Laine, “Elucidating the Design Space of Diffusion-Based Generative Models,” NeurIPS, 2022. arXiv
  21. C. Lu, Y. Zhou, F. Bao, J. Chen, C. Li and J. Zhu, “DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps,” NeurIPS, 2022. arXiv
  22. T. Salimans and J. Ho, “Progressive Distillation for Fast Sampling of Diffusion Models,” ICLR, 2022. arXiv
  23. M. Heusel, H. Ramsauer, T. Unterthiner, B. Nessler and S. Hochreiter, “GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium,” NeurIPS (NIPS), 2017. arXiv
  24. T. Kynkäänniemi, T. Karras, S. Laine, J. Lehtinen and T. Aila, “Improved Precision and Recall Metric for Assessing Generative Models,” NeurIPS, 2019. arXiv
  25. G. Stein, J. C. Cresswell, R. Hosseinzadeh et al., “Exposing Flaws of Generative Model Evaluation Metrics and Their Unfair Treatment of Diffusion Models,” NeurIPS, 2023. arXiv
  26. S. Jayasumana, S. Ramalingam, A. Veit, D. Glasner, A. Chakrabarti and S. Kumar, “Rethinking FID: Towards a Better Evaluation Metric for Image Generation,” CVPR, 2024. CVF
  27. F.-A. Croitoru, V. Hondru, R. T. Ionescu and M. Shah, “Diffusion Models in Vision: A Survey,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023. doi
  28. H. Cao, C. Tan, Z. Gao, Y. Xu, G. Chen, P.-A. Heng and S. Z. Li, “A Survey on Generative Diffusion Models,” IEEE Transactions on Knowledge and Data Engineering, vol. 36, no. 7, pp. 2814–2830, 2024. doi
  29. C.-H. Lai, Y. Song, D. Kim, Y. Mitsufuji and S. Ermon, “The Principles of Diffusion Models,” monograph, arXiv:2510.21890, 2025. arXiv
  30. H. Shen, J. Zhang, B. Xiong et al., “Efficient Diffusion Models: A Survey,” Transactions on Machine Learning Research, 2025. arXiv
  31. A. Melnik, M. Ljubljanac, C. Lu, Q. Yan, W. Ren and H. Ritter, “Video Diffusion Models: A Survey,” Transactions on Machine Learning Research, 2024. arXiv
  32. C. Lu and Y. Song, “Simplifying, Stabilizing and Scaling Continuous-Time Consistency Models,” ICLR, 2025. arXiv
  33. A. Sauer, D. Lorenz, A. Blattmann and R. Rombach, “Adversarial Diffusion Distillation,” ECCV, 2024. arXiv
  34. M. S. Albergo and E. Vanden-Eijnden, “Building Normalizing Flows with Stochastic Interpolants,” ICLR, 2023. arXiv
  35. J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li and L. Fei-Fei, “ImageNet: A Large-Scale Hierarchical Image Database,” CVPR, 2009, pp. 248–255. doi
  36. C. Szegedy, V. Vanhoucke, S. Ioffe, J. Shlens and Z. Wojna, “Rethinking the Inception Architecture for Computer Vision,” CVPR, 2016, pp. 2818–2826. arXiv
  37. O. Ronneberger, P. Fischer and T. Brox, “U-Net: Convolutional Networks for Biomedical Image Segmentation,” MICCAI, LNCS 9351, pp. 234–241, 2015. arXiv
  38. A. Radford, J. W. Kim, C. Hallacy et al., “Learning Transferable Visual Models From Natural Language Supervision,” ICML, 2021. arXiv
  39. M. Oquab, T. Darcet, T. Moutakanni et al., “DINOv2: Learning Robust Visual Features without Supervision,” Transactions on Machine Learning Research, 2024. arXiv
  40. B. Efron, “Tweedie’s Formula and Selection Bias,” Journal of the American Statistical Association, vol. 106, no. 496, pp. 1602–1614, 2011. doi
  41. A. Dosovitskiy, L. Beyer, A. Kolesnikov et al., “An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale,” ICLR, 2021. arXiv
  42. T. Li and K. He, “Back to Basics: Let Denoising Generative Models Denoise,” CVPR, 2026 (preprint 2025). paper, publication record, official JiT code.
  43. B. Zheng, N. Ma, S. Tong and S. Xie, “Diffusion Transformers with Representation Autoencoders,” ICLR, 2026 (preprint 2025). proceedings, manuscript.
  44. S. Tong, B. Zheng, Z. Wang, B. Tang, N. Ma, E. Brown, J. Yang, R. Fergus, Y. LeCun and S. Xie, “Scaling Text-to-Image Diffusion Transformers with Representation Autoencoders,” CVPR Findings Track, 2026. manuscript, author publication record.
  45. D. Podell, Z. English, K. Lacey, A. Blattmann, T. Dockhorn, J. Müller, J. Penna and R. Rombach, “SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis,” ICLR, 2024. proceedings, official model card.
  46. Black Forest Labs, “FLUX.1 [dev],” official pretrained-model card, 2024. model card.
  47. W. Kong, Q. Tian, Z. Zhang et al., “HunyuanVideo: A Systematic Framework For Large Video Generative Models,” technical report, arXiv:2412.03603, 2024. report, official release.
  48. Team Wan, A. Wang, B. Ai, B. Wen et al., “Wan: Open and Advanced Large-Scale Video Generative Models,” technical report, arXiv:2503.20314, 2025. report, official Wan2.1 release.
  49. Z. Geng, Y. Lu, Z. Wu, E. Shechtman, J. Z. Kolter and K. He, “Improved Mean Flows: On the Challenges of Fastforward Generative Models,” CVPR, pp. 30467–30476, 2026. CVF
  50. Y. Lu, S. Lu, Q. Sun, H. Zhao, Z. Jiang, X. Wang, T. Li, Z. Geng and K. He, “One-step Latent-free Image Generation with Pixel Mean Flows,” ICML, PMLR 306, pp. 81265–81277, 2026. PMLR