mlsanityData to 14 Oct 2026

DeCo: Frequency-Decoupled Pixel Diffusion for End-to-End Image Generation

arXiv 2511.19365 ↗First posted 24 Nov 2025cs.CV cs.AIcited by 60

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

Pixel diffusion aims to generate images directly in pixel space in an end-to-end fashion. This approach avoids the limitations of VAE in the two-stage latent diffusion, offering higher model capacity. Existing pixel diffusion models suffer from slow training and inference, as they usually model both high-frequency signals and low-frequency semantics within a single diffusion transformer (DiT). To pursue a more efficient pixel diffusion paradigm, we propose the frequency-DeCoupled pixel diffusion framework. With the intuition to decouple the generation of high and low frequency components, we leverage a lightweight pixel decoder to generate high-frequency details conditioned on semantic guidance from the DiT. This thus frees the DiT to specialize in modeling low-frequency semantics. In addition, we introduce a frequency-aware flow-matching loss that emphasizes visually salient frequencies while suppressing insignificant ones. Extensive experiments show that DeCo achieves superior performance among pixel diffusion models, attaining FID of 1.62 (256x256) and 2.22 (512x512) on ImageNet, closing the gap with latent diffusion methods. Furthermore, our pretrained text-to-image model achieves a leading overall score of 0.86 on GenEval in system-level comparison. Codes are publicly available at https://github.com/Zehong-Ma/DeCo.

Key facts. Cited by 60 papers in the map (data through 14 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 69 of its 81 references that are in the map. First posted on arXiv 24 Nov 2025.

Builds on

69 of 81 references.

Closest ancestors · 26

Same direction, newest first

  1. 2025PixNerd: Pixel Neural Field Diffusion31 Jul
  2. 2025REPA-E: Unlocking VAE for End-to-End Tuning with Latent Diffusion Transformers14 Apr
  3. 2025U-REPA: Aligning Diffusion U-Nets to ViTs24 Mar
  4. 2025Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models2 Jan
  5. 2024DiC: Rethinking Conv3x3 Designs in Diffusion Models31 Dec
  6. 2024Deep Compression Autoencoder for Efficient High-Resolution Diffusion Models14 Oct
  7. 2024Diffusion Models Need Visual Priors for Image Generation11 Oct
  8. 2024Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Think9 Oct
  9. 2024U-DiTs: Downsample Tokens in U-Shaped Diffusion Transformers4 May
  10. 2024Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Models11 Apr
  11. 2024Score identity Distillation: Exponentially Fast Distillation of Pretrained Diffusion Models for One-Step Generation5 Apr
  12. 2024Scaling Rectified Flow Transformers for High-Resolution Image Synthesis5 Mar
14 more
  1. 2024SiT: Exploring Flow and Diffusion-based Generative Models with Scalable Interpolant Transformers16 Jan
  2. 2023Analyzing and Improving the Training Dynamics of Diffusion Models5 Dec
  3. 2023Latent Consistency Models: Synthesizing High-Resolution Images with Few-Step Inference6 Oct
  4. 2023PixArt-α: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis30 Sep
  5. 2023Relay Diffusion: Unifying diffusion process across resolutions for image synthesis4 Sep
  6. 2023Understanding Diffusion Objectives as the ELBO with Simple Data Augmentation1 Mar
  7. 2023Simple diffusion: End-to-end diffusion for high resolution images26 Jan
  8. 2022Scalable Diffusion Models with Transformers19 Dec
  9. 2022All are Worth Words: A ViT Backbone for Diffusion Models25 Sep
  10. 2022Classifier-Free Diffusion Guidance26 Jul
  11. 2021High-Resolution Image Synthesis with Latent Diffusion Models20 Dec
  12. 2021Diffusion Models Beat GANs on Image Synthesis11 May
  13. 2020Denoising Diffusion Implicit Models6 Oct
  14. 2020Denoising Diffusion Probabilistic Models19 Jun

Other references · 43

  1. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,713
  2. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 14,132
  3. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 13,028
  4. 2025Qwen3 Technical Reportcited by 8,820
  5. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,974
  6. 2022Flow Matching for Generative Modelingcited by 5,702
  7. 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,703
  8. 2017Progressive Growing of GANs for Improved Quality, Stability, and Variationcited by 2,988
  9. 2016Improved Techniques for Training GANscited by 2,970
  10. 2023GenEval: An Object-Focused Framework for Evaluating Text-to-Image Alignmentcited by 972
  11. 2017Sigmoid-Weighted Linear Units for Neural Network Function Approximation in Reinforcement Learningcited by 950
  12. 2019Improved Precision and Recall Metric for Assessing Generative Modelscited by 944
31 more
  1. 2022Vision Transformer Adapter for Dense Predictionscited by 565
  2. 2024ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignmentcited by 475
  3. 2025BLIP3-o: A Family of Fully Open Unified Multimodal Models-Architecture, Training and Datasetcited by 376
  4. 2019High Frequency Component Helps Explain the Generalization of Convolutional Neural Networkscited by 310
  5. 2022How Do Vision Transformers Work?cited by 307
  6. 2021Generating Images with Sparse Representationscited by 266
  7. 2025OmniGen2: Towards Instruction-Aligned Multimodal Generationcited by 253
  8. 2015Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networkscited by 214
  9. 2022Scalable Adaptive Computation for Iterative Generationcited by 195
  10. 2025Seedream 3.0 Technical Reportcited by 154
  11. 2024Fluid: Scaling Autoregressive Text-to-image Generative Models with Continuous Tokenscited by 147
  12. 2025Mogao: An Omni Foundation Model for Interleaved Multi-Modal Generationcited by 109
  13. 2022Inception Transformercited by 99
  14. 2025PixelFlow: Pixel-Space Generative Models with Flowcited by 97
  15. 2024Normalizing Flows are Capable Generative Modelscited by 91
  16. 2024DiM: Diffusion Mamba for Efficient High-Resolution Image Synthesiscited by 72
  17. 2025Seedream 2.0: A Native Chinese-English Bilingual Image Generation Foundation Modelcited by 62
  18. 2024CustomCrafter: Customized Video Generation with Preserving Motion and Concept Composition Abilitiescited by 60
  19. 2024JetFormer: An Autoregressive Generative Model of Raw Images and Textcited by 46
  20. 2025Fractal Generative Modelscited by 46
  21. 2025MagCache: Fast Video Generation with Magnitude-Aware Cachecited by 41
  22. 2024MV-VTON: Multi-View Virtual Try-On with Diffusion Modelscited by 25
  23. 2022FreGAN: Exploiting Frequency Components for Training GANs under Limited Datacited by 16
  24. 2024VideoMaker: Zero-shot Customized Video Generation with the Inherent Force of Video Diffusion Modelscited by 16
  25. 2025Native-Resolution Image Synthesiscited by 15
  26. 2025Frequency-Aware Autoregressive Modeling for Efficient High-Resolution Image Synthesiscited by 15
  27. 2024SphereDiffusion: Spherical Geometry-Aware Distortion Resilient Diffusion Modelcited by 13
  28. 2021Exploring The Effect of High-frequency Components in GANs Trainingcited by 10
  29. 2025DMM: Building a Versatile Image Generation Model via Distillation-Based Model Mergingcited by 5
  30. 2025MoSA: Motion-Coherent Human Video Generation via Structure-Appearance Decouplingcited by 4
  31. 2025MultiCrafter: High-Fidelity Multi-Subject Generation via Disentangled Attention and Identity-Aware Preference Alignmentcited by 4

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 3; 2026: 57.

60 papers.

Sort:

By area:

  1. 2025What matters for Representation Alignment: Global Information or Spatial Structure?cited by 84
  2. 2026One-step Latent-free Image Generation with Pixel Mean Flowscited by 53
  3. 2026PixelGen: Improving Pixel Diffusion with Perceptual Supervisioncited by 14
  4. 2026L2P: Unlocking Latent Potential for Pixel Generationcited by 9
  5. 2026HyperDiT: Hyper-Connected Transformers for High-Fidelity Pixel-Space Diffusioncited by 9
  6. 2025EMMA: Efficient Multimodal Understanding, Generation, and Editing with a Unified Architecturecited by 8
  7. 2026Asymmetric Flow Modelscited by 8
  8. 2026PixelU: A U-Shaped Transformer for Efficient End-to-End Pixel Diffusioncited by 8
  9. 2026CoD-Lite: Real-Time Diffusion-Based Generative Image Compressioncited by 7
  10. 2026An Empirical Study of Training Pixel-Space Text-to-Image Diffusion Modelscited by 6
  11. 2026FrequencyBooster: Full-Frequency Modeling for High-Fidelity Pixel Diffusioncited by 4
  12. 2026PixWorld: Unifying 3D Scene Generation and Reconstruction in Pixel Spacecited by 4
38 more
  1. 2026FREPix: Frequency-Heterogeneous Flow Matching for Pixel-Space Image Generationcited by 3
  2. 2026Why Do DiT Editors Drift? Plug-and-Play Low Frequency Alignment in VAE Latent Spacecited by 3
  3. 2026V-Co: A Closer Look at Visual Representation Alignment via Co-Denoisingcited by 2
  4. 2026Trajectory Forcing: Structure-First Generation with Controllable Semantic Trajectoriescited by 2
  5. 2026PixGS: Pixel-Space Diffusion for Direct 3D Gaussian Splat Generationcited by 2
  6. 2026SPEED: One-Step Pixel Diffusion for High-quality Video Frame Interpolationcited by 2
  7. 2026A Frozen Pixel-Space Diffusion Model Can Guide Itself with Its Own Samplescited by 2
  8. 2026GenFirst: Generation Before Reconstruction for Stable End-to-End Latent Generative Modelingcited by 2
  9. 2026Persistence Forcing: Exploiting Feature Specialization in Pixel-Space Diffusioncited by 2
  10. 2025RePack then Refine: Efficient Diffusion Transformer with Vision Foundation Modelcited by 1
  11. 2026WiT: Waypoint Diffusion Transformers for Alleviating Trajectory Conflict in Pixel-Space Image Generationcited by 1
  12. 2026SPROUT: A Scalable Diffusion Foundation Model for Multi-Crop Plant Phenotypingcited by 1
  13. 2026PixIE: Prompted Pixel-Space Low-Light Image Enhancementcited by 1
  14. 2026WavTTS: Towards High-Quality Zero-Shot TTS via Direct Raw Waveform Modelingcited by 1
  15. 2026BareWave: Waveform-Native Flow-Matching Text-to-Speechcited by 1
  16. 2026Pixel-Space Diffusion Transformerscited by 1
  17. 2026DuSPiT: Dual-Branch Sub-Patch Pixel Diffusion Transformercited by 1
  18. 2026Generative Video Compression with Adaptive Score Distillationcited by 1
  19. 2026PixelControl: Fine-Grained Condition Fidelity in Text-to-Image Diffusioncited by 1
  20. 2026PXDepth: Pixel-Space Modeling for Structure Preserving Monocular Depth Estimationcited by 1
  21. 2026Pixel-Space Diffusion via Observation Operatorscited by 1
  22. 2026Balancing Frequencies and Pixels in Flow Matchingcited by 1
  23. 2026Coevolving Representations in Joint Image-Feature Diffusioncited by 0
  24. 2026UniCSG: Unified High-Fidelity Content-Constrained Style-Driven Generation via Staged Semantic and Frequency Disentanglementcited by 0
  25. 2026HDRFace: Rethinking Face Restoration with High-Dimensional Representationcited by 0
  26. 2026RaPD: Resolution-Agnostic Pixel Diffusion via Semantics-Enriched Implicit Representationscited by 0
  27. 2026Diffusion Image Generation with Explicit Modeling of Data Manifold Geometrycited by 0
  28. 2026SpectralDiT: Timestep-Conditioned Spectral Residual Correction for Flow-Matching DiTscited by 0
  29. 2026FADRA: Frequency-Aware Diffusion with Residual Adaptation for Video Face Restorationcited by 0
  30. 2026STEREOFLOW: Progressive Stereo Matching with StereoDiT and Transition Flow Matchingcited by 0
  31. 2026MOSAIK: Multi-Patch Content-Aware Spatial Allocation of Image Tokens for Efficient Generationcited by 0
  32. 2026Energy-Guided Flow Matchingcited by 0
  33. 2026Advanced Pixel Diffusion Model with Guided Sparse Global Refinementcited by 0
  34. 2026PixelDiT2: Representation-Grounded Pixel Diffusion Transformerscited by 0
  35. 2026Spectral Amplitude Purification in Distribution Matching for Diffusion Distillationcited by 0
  36. 2026Residual-Stream Burden Shapes Representation Learning in Diffusion Transformerscited by 0
  37. 2026LDM-is-AE: Latent Diffusion Model is an Auto-Encoder for End-to-End Image Generationcited by 0
  38. 2026DMA2: Pixel-space Distribution Matching with Adversarial and Anchor Lossescited by 0

50 most cited of 60 All 60 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025DiP: Taming Diffusion Models in Pixel Space
  2. 2025PixelDiT: Pixel Diffusion Transformers for Image Generation
  3. 2026PiD: Fast and High-Resolution Latent Decoding with Pixel Diffusion
  4. 2024Simpler Diffusion (SiD2): 1.5 FID on ImageNet512 with pixel-space diffusion
  5. 2024FAM Diffusion: Frequency and Attention Modulation for High-Resolution Image Generation with Stable Diffusion
  6. 2025FreSca: Scaling in Frequency Space Enhances Diffusion Models
  7. 2023Faster Diffusion: Rethinking the Role of the Encoder for Diffusion Model Inference
  8. 2025Toward Lightweight and Fast Decoders for Diffusion Models in Image and Video Generation