mlsanityData to 14 Oct 2026

SRC-Flow: Compact Semantic Representations Enable Normalizing Flows for Image Generation

arXiv 2605.18267 ↗First posted 18 May 2026cs.CVcited by 0

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

Abstract (arXiv)

Normalizing flows (NFs) provide exact likelihoods and deterministic invertible sampling, but have historically lagged behind diffusion models for large-scale image generation. We identify a key obstacle: NFs are required to learn a single invertible transport over the full ambient space, making them highly sensitive to high-dimensional representations. This leads to a semantic-capacity mismatch in modern visual representation spaces, where semantic information is compact but encoded in overcomplete features. We propose SRC-Flow, which introduces a Semantic Representation Compressor (SRC) to compact high-dimensional RAE features into a low-dimensional semantic space before flow modeling and preserve reconstruction through the frozen RAE decoder. This compact space reduces the modeling burden of NFs and enables effective likelihood-based generation in semantic representation space. We further adopt constant noise regularization tailored to the fixed unconditional bijection learned by flows. On ImageNet 256x256 and 512x512, SRC-Flow achieves state-of-the-art generation quality among normalizing flow methods, with gFID scores of 1.65 and 2.07 under classifier-free guidance, while retaining exact likelihood computation in the compact semantic representation space and deterministic invertible sampling at the flow level. Codes and models will be released in the future.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Flow matching, time series and science, one of 14 directions in Diffusion models. Builds on 67 of its 72 references that are in the map. First posted on arXiv 18 May 2026.

Builds on

67 of 72 references.

Closest ancestors · 3

Same direction, newest first

  1. 2026Normalizing Flows with Iterative Denoising21 Apr
  2. 2022Flow Matching for Generative Modeling6 Oct
  3. 2022Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow7 Sep

Other references · 64

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  3. 2020Denoising Diffusion Probabilistic Modelscited by 21,076
  4. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  5. 2020Denoising Diffusion Implicit Modelscited by 9,781
  6. 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 9,122
  7. 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,255
  8. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,974
  9. 2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,403
  10. 2021Emerging Properties in Self-Supervised Vision Transformerscited by 6,968
  11. 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,585
  12. 2022Scalable Diffusion Models with Transformerscited by 6,054
38 more
  1. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesiscited by 4,356
  2. 2021Zero-Shot Text-to-Image Generationcited by 4,288
  3. 2024Scaling Rectified Flow Transformers for High-Resolution Image Synthesiscited by 4,181
  4. 2018Neural Ordinary Differential Equationscited by 3,779
  5. 2020Taming Transformers for High-Resolution Image Synthesiscited by 3,253
  6. 2022Elucidating the Design Space of Diffusion-Based Generative Modelscited by 3,045
  7. 2016Improved Techniques for Training GANscited by 2,970
  8. 2018Large Scale GAN Training for High Fidelity Natural Image Synthesiscited by 2,626
  9. 2015Variational Inference with Normalizing Flowscited by 2,170
  10. 2016Density estimation using Real NVPcited by 2,042
  11. 2018Glow: Generative Flow with Invertible 1x1 Convolutionscited by 1,545
  12. 2025SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Featurescited by 1,188
  13. 2014NICE: Non-linear Independent Components Estimationcited by 1,058
  14. 2024Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Predictioncited by 1,023
  15. 2019Improved Precision and Recall Metric for Assessing Generative Modelscited by 944
  16. 2016Pixel Recurrent Neural Networkscited by 816
  17. 2024SiT: Exploring Flow and Diffusion-based Generative Models with Scalable Interpolant Transformerscited by 737
  18. 2024Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Thinkcited by 703
  19. 2017Masked Autoregressive Flow for Density Estimationcited by 684
  20. 2024Autoregressive Image Generation without Vector Quantizationcited by 651
  21. 2016Improving Variational Inference with Inverse Autoregressive Flowcited by 640
  22. 2023Language Model Beats Diffusion -- Tokenizer is Key to Visual Generationcited by 640
  23. 2019Neural Spline Flowscited by 596
  24. 2018FFJORD: Free-form Continuous Dynamics for Scalable Reversible Generative Modelscited by 594
  25. 2022StyleGAN-XL: Scaling StyleGAN to Large Diverse Datasetscited by 492
  26. 2023Analyzing and Improving the Training Dynamics of Diffusion Modelscited by 402
  27. 2025Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Modelscited by 393
  28. 2018Invertible Residual Networkscited by 365
  29. 2025Diffusion Transformers with Representation Autoencoderscited by 295
  30. 2019Flow++: Improving Flow-Based Generative Models with Variational Dequantization and Architecture Designcited by 262
  31. 2023MDTv2: Masked Diffusion Transformer is a Strong Image Synthesizercited by 262
  32. 2019Residual Flows for Invertible Generative Modelingcited by 257
  33. 2025REPA-E: Unlocking VAE for End-to-End Tuning with Latent Diffusion Transformerscited by 225
  34. 2022Scalable Adaptive Computation for Iterative Generationcited by 195
  35. 2023Fast Training of Diffusion Models with Masked Transformerscited by 176
  36. 2025DDT: Decoupled Diffusion Transformercited by 109
  37. 2023DiffiT: Diffusion Vision Transformers for Image Generationcited by 106
  38. 2025PixelFlow: Pixel-Space Generative Models with Flowcited by 97

50 most cited of 64 All 64 (JSON)

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026FlowSSC: Universal Generative Monocular Semantic Scene Completion via One-Step Latent Diffusion
  2. 2024Jet: A Modern Transformer-Based Normalizing Flow
  3. 2025RecTok: Reconstruction Distillation along Rectified Flow
  4. 2026On the Diffusibility of High-Dimensional Latents
  5. 2026FlowCodec: One-Step Flow Prior for Generative Image Compression
  6. 2025REGLUE Your Latents with Global and Local Semantics for Entangled Diffusion
  7. 2026InstantFusion: A Shared Latent Interface for Heterogeneous Image Generators
  8. 2024JanusFlow: Harmonizing Autoregression and Rectified Flow for Unified Multimodal Understanding and Generation