mlsanityData to 14 Oct 2026

AlphaFlow: Understanding and Improving MeanFlow Models

arXiv 2510.20771 ↗First posted 23 Oct 2025cs.CV cs.LGcited by 41

On the map: Diffusion models → Flow matching, time series and science

Abstract (arXiv)

MeanFlow has recently emerged as a powerful framework for few-step generative modeling trained from scratch, but its success is not yet fully understood. In this work, we show that the MeanFlow objective naturally decomposes into two parts: trajectory flow matching and trajectory consistency. Through gradient analysis, we find that these terms are strongly negatively correlated, causing optimization conflict and slow convergence. Motivated by these insights, we introduce α-Flow, a broad family of objectives that unifies trajectory flow matching, Shortcut Model, and MeanFlow under one formulation. By adopting a curriculum strategy that smoothly anneals from trajectory flow matching to MeanFlow, α-Flow disentangles the conflicting objectives, and achieves better convergence. When trained from scratch on class-conditional ImageNet-1K 256x256 with vanilla DiT backbones, α-Flow consistently outperforms MeanFlow across scales and settings. Our largest α-Flow-XL/2+ model achieves new state-of-the-art results using vanilla DiT backbones, with FID scores of 2.58 (1-NFE) and 2.15 (2-NFE).

Key facts. Cited by 41 papers in the map (data through 14 Oct 2026). Placed in Flow matching, time series and science, one of 14 directions in Diffusion models. Builds on 41 of its 45 references that are in the map. First posted on arXiv 23 Oct 2025.

Builds on

41 of 45 references.

Closest ancestors · 5

Same direction, newest first

  1. 2025SplitMeanFlow: Interval Splitting Consistency in Few-Step Generative Modeling22 Jul
  2. 2025FACM: Flow-Anchored Consistency Models4 Jul
  3. 2024Consistency Flow Matching: Defining Straight Flows with Velocity Consistency2 Jul
  4. 2022Flow Matching for Generative Modeling6 Oct
  5. 2022Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow7 Sep

Other references · 36

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2020Denoising Diffusion Probabilistic Modelscited by 21,076
  3. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  4. 2020Denoising Diffusion Implicit Modelscited by 9,781
  5. 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 9,122
  6. 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,255
  7. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,974
  8. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,058
  9. 2022Scalable Diffusion Models with Transformerscited by 6,054
  10. 2022Classifier-Free Diffusion Guidancecited by 5,847
  11. 2024Scaling Rectified Flow Transformers for High-Resolution Image Synthesiscited by 4,181
  12. 2019Generative Modeling by Estimating Gradients of the Data Distributioncited by 4,101
24 more
  1. 2022Elucidating the Design Space of Diffusion-Based Generative Modelscited by 3,045
  2. 2023Consistency Modelscited by 1,822
  3. 2023One-step Diffusion with Distribution Matching Distillationcited by 1,019
  4. 2022Building Normalizing Flows with Stochastic Interpolantscited by 995
  5. 2024WorldSimBench: Towards Video Generation Models as World Simulatorscited by 881
  6. 2023Adversarial Diffusion Distillationcited by 759
  7. 2024Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Thinkcited by 703
  8. 2024Improved Distribution Matching Distillation for Fast Image Synthesiscited by 684
  9. 2025Mean Flows for One-step Generative Modelingcited by 561
  10. 2023Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusioncited by 459
  11. 2023Improved Techniques for Training Consistency Modelscited by 439
  12. 2023Analyzing and Improving the Training Dynamics of Diffusion Modelscited by 402
  13. 2024One Step Diffusion via Shortcut Modelscited by 391
  14. 2024Simplifying, Stabilizing and Scaling Continuous-Time Consistency Modelscited by 283
  15. 2018An Empirical Model of Large-Batch Trainingcited by 281
  16. 2024Score identity Distillation: Exponentially Fast Distillation of Pretrained Diffusion Models for One-Step Generationcited by 214
  17. 2022The Role of ImageNet Classes in Fr'echet Inception Distancecited by 213
  18. 2024Consistency Models Made Easycited by 144
  19. 2023TRACT: Denoising Diffusion Models with Transitive Closure Time-Distillationcited by 116
  20. 2025Inductive Moment Matchingcited by 113
  21. 2024Phased Consistency Modelscited by 85
  22. 2025Small Batch Size Training for Language Models: When Vanilla SGD Works, and Why Gradient Accumulation Is Wastefulcited by 41
  23. 2025Transition Models: Rethinking the Generative Learning Objectivecited by 34
  24. 2025Unified Continuous Generative Modelscited by 19

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 8; 2026: 33.

41 papers.

Sort:

By area:

  1. 2025Improved Mean Flows: On the Challenges of Fastforward Generative Modelscited by 110
  2. 2026One-Step Generative Modeling via Wasserstein Gradient Flowscited by 24
  3. 2026SnapFlow: One-Step Action Generation for Flow-Matching VLAs via Progressive Self-Distillationcited by 14
  4. 2025Adversarial Flow Modelscited by 12
  5. 2025Learning Straight Flows: Variational Flow Matching for Efficient Generationcited by 8
  6. 2025From Navigation to Refinement: Revealing the Two-Stage Nature of Flow-based Diffusion Models through Oracle Velocitycited by 8
  7. 2025Bidirectional Normalizing Flow: From Data to Noise and Backcited by 7
  8. 2026Learning Hamiltonian Flow Maps: Mean Flow Consistency for Large-Timestep Molecular Dynamicscited by 7
  9. 2025MeanFlow Transformers with Representation Autoencoderscited by 6
  10. 2026Riemannian MeanFlow for One-Step Generation on Manifoldscited by 5
  11. 2025Understanding, Accelerating, and Improving MeanFlow Trainingcited by 4
  12. 2025MeanFlow-TSE: One-Step Generative Target Speaker Extraction with Mean Flowcited by 4
29 more
  1. 2026SubFlow: Sub-mode Conditioned Flow Matching for Diverse One-Step Generationcited by 4
  2. 2026Bi-Anchor Interpolation Solver for Accelerating Generative Modelingcited by 3
  3. 2026EchoLVFM: One-Step Video Generation via Latent Flow Matching for Echocardiogram Synthesiscited by 3
  4. 2026Tyche: One Step Flow for Efficient Probabilistic Weather Forecastingcited by 3
  5. 2026Trajectory Consistency for One-Step Generation on Euler Mean Flowscited by 2
  6. 2026Transition Flow Matchingcited by 2
  7. 2026FlowInOne:Unifying Multimodal Generation as Image-in, Image-out Flow Matchingcited by 2
  8. 2026Extending One-Step Image Generation from Class Labels to Text via Discriminative Text Representationcited by 2
  9. 2026Drift Flow Matchingcited by 2
  10. 2026Trajectory Forcing: Structure-First Generation with Controllable Semantic Trajectoriescited by 2
  11. 2026Difficulty-Calibrated Interpolation Paths for Conditional Flow Matchingcited by 2
  12. 2026WFR-MFM: One-Step Inference for Dynamic Unbalanced Optimal Transportcited by 1
  13. 2026Efficient Hybrid SE(3)-Equivariant Visuomotor Flow Policy via Spherical Harmonics for Robot Manipulationcited by 1
  14. 2026Efficient Image Synthesis with Sphere Latent Encodercited by 1
  15. 2026Improved Distributional Diffusion Modelscited by 1
  16. 2026FlowConsist: Make Your Flow Consistent with Real Trajectorycited by 0
  17. 2026Dual-End Consistency Modelcited by 0
  18. 2026AlphaFlowTSE: One-Step Generative Target Speaker Extraction via Conditional AlphaFlowcited by 0
  19. 2026Point-MF: Stabilizing One-Step Mean Flows for Single-Image Point Cloud Reconstructioncited by 0
  20. 2026Autoregressive One-Step Generative Modeling for Dynamical System Forecastingcited by 0
  21. 2026Accelerating Redshift-Conditioned Galaxy Image Synthesis with One-step Generative Modelingcited by 0
  22. 2026Cross-scale Aligned Supervision for Training GANscited by 0
  23. 2026SE(3)-MeanFlow: Few-Step Protein Backbone Generation on Lie Groupscited by 0
  24. 2026Rethinking Pixel Mean Flows via Interval Denoisercited by 0
  25. 2026FlashRender: Few-Step Generative Rendering via Camera-Controlled Video MeanFlowcited by 0
  26. 2026A Latent Distribution Perspective on Evaluating and Improving Latent Generative Modelscited by 0
  27. 2026Improving Generative Model Self-Training with Geometrically Modified Outputscited by 0
  28. 2026One-Step Generative Modeling via Training Dynamics Actioncited by 0
  29. 2026Kinematic MeanFlow: One-Step Action Generation Policy for Robotic Foundation Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Modular MeanFlow: Towards Stable and Scalable One-Step Generative Modeling
  2. 2025Overcoming the Curvature Bottleneck in MeanFlow
  3. 2023Flow Matching in Latent Space
  4. 2025Towards High-Order Mean Flow Generative Models: Feasibility, Expressivity, and Provably Efficient Criteria
  5. 2026XYZFlow:Scaling Multi dimensional Shortcut Flows for Efficient Generative Modeling
  6. 2025Theoretical Guarantees for High Order Trajectory Refinement in Generative Flows
  7. 2026MintFlow: Minimal Trajectory Intervention for Constrained Flow Matching
  8. 2026Mean Flow Distillation: Robust and Stable Distillation for Flow Matching Models