mlsanityData to 7 Oct 2026

Improved Mean Flows: On the Challenges of Fastforward Generative Models

arXiv 2512.02012 ↗First posted 1 Dec 2025cs.CV cs.LGcited by 102

Not on the map. Why?

Cited by 45 in Diffusion models, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

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 v, re-parameterized by a network that predicts the average velocity u. 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 improved MeanFlow (iMF) method, trained entirely from scratch, achieves 1.72 FID with a single function evaluation (1-NFE) on ImageNet 256×256. 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.

Key facts. Cited by 102 papers in the map (data through 7 Oct 2026). Builds on 55 of its 61 references that are in the map. First posted on arXiv 1 Dec 2025.

Builds on

55 of 61 references.

References · 55

  1. 2017Attention Is All You Needcited by 59,932
  2. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  3. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  4. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
  5. 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,185
  6. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,005
  7. 2022Scalable Diffusion Models with Transformerscited by 5,931
  8. 2022Classifier-Free Diffusion Guidancecited by 5,717
  9. 2022Flow Matching for Generative Modelingcited by 5,426
  10. 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,581
  11. 2019Generative Modeling by Estimating Gradients of the Data Distributioncited by 4,067
  12. 2022Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flowcited by 3,550
38 more
  1. 2022Elucidating the Design Space of Diffusion-Based Generative Modelscited by 2,994
  2. 2016Improved Techniques for Training GANscited by 2,938
  3. 2018Large Scale GAN Training for High Fidelity Natural Image Synthesiscited by 2,609
  4. 2023Consistency Modelscited by 1,778
  5. 2020GLU Variants Improve Transformercited by 1,724
  6. 2017Accurate, Large Minibatch SGD: Training ImageNet in 1 Hourcited by 1,479
  7. 2019Root Mean Square Layer Normalizationcited by 1,425
  8. 2022MaskGIT: Masked Generative Image Transformercited by 1,082
  9. 2024Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Predictioncited by 1,005
  10. 2022Building Normalizing Flows with Stochastic Interpolantscited by 964
  11. 2023Stochastic Interpolants: A Unifying Framework for Flows and Diffusionscited by 897
  12. 2024Autoregressive Model Beats Diffusion: Llama for Scalable Image Generationcited by 760
  13. 2022On Distillation of Guided Diffusion Modelscited by 721
  14. 2024SiT: Exploring Flow and Diffusion-based Generative Models with Scalable Interpolant Transformerscited by 707
  15. 2022Autoregressive Image Generation using Residual Quantizationcited by 706
  16. 2024Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Thinkcited by 666
  17. 2024Autoregressive Image Generation without Vector Quantizationcited by 644
  18. 2025Mean Flows for One-step Generative Modelingcited by 524
  19. 2023Scaling up GANs for Text-to-Image Synthesiscited by 492
  20. 2022StyleGAN-XL: Scaling StyleGAN to Large Diverse Datasetscited by 489
  21. 2023Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusioncited by 448
  22. 2023Improved Techniques for Training Consistency Modelscited by 430
  23. 2023Simple diffusion: End-to-end diffusion for high resolution imagescited by 413
  24. 2025Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Modelscited by 376
  25. 2024One Step Diffusion via Shortcut Modelscited by 373
  26. 2024Simplifying, Stabilizing and Scaling Continuous-Time Consistency Modelscited by 276
  27. 2025Diffusion Transformers with Representation Autoencoderscited by 275
  28. 2024Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Modelscited by 240
  29. 2024Consistency Models Made Easycited by 140
  30. 2024Randomized Autoregressive Visual Generationcited by 125
  31. 2025Inductive Moment Matchingcited by 107
  32. 2025DDT: Decoupled Diffusion Transformercited by 100
  33. 2024Simpler Diffusion (SiD2): 1.5 FID on ImageNet512 with pixel-space diffusioncited by 74
  34. 2025Beyond Next-Token: Next-X Prediction for Autoregressive Visual Generationcited by 71
  35. 2024JetFormer: An Autoregressive Generative Model of Raw Images and Textcited by 44
  36. 2025AlphaFlow: Understanding and Improving MeanFlow Modelscited by 40
  37. 2025Transition Models: Rethinking the Generative Learning Objectivecited by 32
  38. 2025STARFlow: Scaling Latent Normalizing Flows for High-resolution Image Synthesiscited by 31

50 most cited of 55 All 55 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 1; 2026: 101.

102 papers.

Sort:

By area:

  1. 2026Generative Modeling via Driftingcited by 98
  2. 2026One-step Latent-free Image Generation with Pixel Mean Flowscited by 49
  3. 2026Categorical Flow Mapscited by 37
  4. 2026Meta Flow Maps enable scalable reward alignmentcited by 32
  5. 2026Representation Fr'echet Loss for Visual Generationcited by 32
  6. 2026ELF: Embedded Language Flowscited by 27
  7. 2026Discrete Flow Mapscited by 23
  8. 2026One-Step Generative Modeling via Wasserstein Gradient Flowscited by 18
  9. 2026PixelGen: Improving Pixel Diffusion with Perceptual Supervisioncited by 10
  10. 2026Representation Distribution Matching for One-Step Visual Generationcited by 10
  11. 2025Bidirectional Normalizing Flow: From Data to Noise and Backcited by 7
  12. 2026Parallel Decoding Distillation for Fast Image and Video Generationcited by 7
38 more
  1. 2026Kernel-Gradient Drifting Modelscited by 6
  2. 2026Learning Hamiltonian Flow Maps: Mean Flow Consistency for Large-Timestep Molecular Dynamicscited by 5
  3. 2026One-Step Flow Policy: Self-Distillation for Fast Visuomotor Policiescited by 5
  4. 2026How to Guide Your Flow: Few-Step Alignment via Flow Map Reward Guidancecited by 5
  5. 2026Drift-Based Policy Optimization: Native One-Step Policy Learning for Online Robot Controlcited by 4
  6. 2026Distribution Matching Distillation without Fake Score Networkcited by 4
  7. 2026Strong Stochastic Flow Mapscited by 4
  8. 2026MeanFlowNFT: Bringing Forward-Process RL to Average-Velocity Generatorscited by 4
  9. 2026GE-Act 2.0: Pretraining and Scaling a World-Action Model for Robotic Manipulationcited by 4
  10. 2026Bi-Anchor Interpolation Solver for Accelerating Generative Modelingcited by 3
  11. 2026Riemannian MeanFlow for One-Step Generation on Manifoldscited by 3
  12. 2026Curriculum Sampling: A Two-Phase Curriculum for Efficient Training of Flow Matchingcited by 3
  13. 2026MENO: MeanFlow-Enhanced Neural Operators for Dynamical Systemscited by 3
  14. 2026Tyche: One Step Flow for Efficient Probabilistic Weather Forecastingcited by 3
  15. 2026Aligning Few-Step Generative Models by Amortizing Sample-based Variational Inferencecited by 3
  16. 2026Mean Flow Distillation: Robust and Stable Distillation for Flow Matching Modelscited by 3
  17. 2026Flow-Map GRPO: Reinforcement Learning for Few-Step Flow-Map Generators via Anchored Stochastic Compositioncited by 3
  18. 2026Stabilizing Consistency Training: A Flow Map Analysis and Self-Distillationcited by 2
  19. 2026Trajectory Consistency for One-Step Generation on Euler Mean Flowscited by 2
  20. 2026Setting-Matched and Semantics-Scaled Benchmarking of One-Step Generative Models Against Multistep Diffusion and Flow Modelscited by 2
  21. 2026One-Shot Generative Flows: Existence and Obstructionscited by 2
  22. 2026Speech Enhancement Based on Drifting Modelscited by 2
  23. 2026CoFlow: Coordinated Few-Step Flow for Offline Multi-Agent Decision Makingcited by 2
  24. 2026Registers Matter for Pixel-Space Diffusion Transformerscited by 2
  25. 2026ReactVLA: Fast and Lightweight Reactive Robot Manipulation via Improved Mean Flow Action Generationcited by 2
  26. 2026Distilling Drifting Transformers with Representation Autoencoderscited by 2
  27. 2026CrossFlow: One-Step Generation Across Latent and Pixel Spacescited by 2
  28. 2026Trajectory Forcing: Structure-First Generation with Controllable Semantic Trajectoriescited by 2
  29. 2026NullFlow: One-Step Generative Reconstructioncited by 2
  30. 2026DiffusionBench: On Holistic Evaluation of Diffusion Transformerscited by 2
  31. 2026Amortized Moment Matching for Visual Generationcited by 2
  32. 2026Accelerating Diffusion Sampling via Speculative Draft Treescited by 2
  33. 2026WFR-MFM: One-Step Inference for Dynamic Unbalanced Optimal Transportcited by 1
  34. 2026Accelerated Sequential Flow Matching: A Bayesian Filtering Perspectivecited by 1
  35. 2026Boltzmann Generators for Condensed Matter via Riemannian Flow Matchingcited by 1
  36. 2026Preconditioned One-Step Generative Modeling for Bayesian Inverse Problems in Function Spacescited by 1
  37. 2026Transition Flow Matchingcited by 1
  38. 2026On Variance Reduction in Learning Mean Flowscited by 1

50 most cited of 102 All 102 (JSON)