mlsanityData to 14 Oct 2026

Attention Distillation: A Unified Approach to Visual Characteristics Transfer

arXiv 2502.20235 ↗First posted 27 Feb 2025cs.CVcited by 33

On the map: Diffusion models → Image generation, text-to-image and editing

Abstract (arXiv)

Recent advances in generative diffusion models have shown a notable inherent understanding of image style and semantics. In this paper, we leverage the self-attention features from pretrained diffusion networks to transfer the visual characteristics from a reference to generated images. Unlike previous work that uses these features as plug-and-play attributes, we propose a novel attention distillation loss calculated between the ideal and current stylization results, based on which we optimize the synthesized image via backpropagation in latent space. Next, we propose an improved Classifier Guidance that integrates attention distillation loss into the denoising sampling process, further accelerating the synthesis and enabling a broad range of image generation applications. Extensive experiments have demonstrated the extraordinary performance of our approach in transferring the examples' style, appearance, and texture to new images in synthesis. Code is available at https://github.com/xugao97/AttentionDistillation.

Key facts. Cited by 33 papers in the map (data through 14 Oct 2026). Placed in Image generation, text-to-image and editing, one of 14 directions in Diffusion models. Builds on 48 of its 73 references that are in the map. First posted on arXiv 27 Feb 2025.

Builds on

48 of 73 references.

Closest ancestors · 16

Same direction, newest first

  1. 2024CSGO: Content-Style Composition in Text-to-Image Generation29 Aug
  2. 2024RB-Modulation: Training-Free Personalization of Diffusion Models using Stochastic Optimal Control27 May
  3. 2024Visual Style Prompting with Swapping Self-Attention20 Feb
  4. 2023Style Injection in Diffusion: A Training-free Approach for Adapting Large-scale Diffusion Models for Style Transfer11 Dec
  5. 2023ControlStyle: Text-Driven Stylized Image Generation Using Diffusion Priors9 Nov
  6. 2023StyleDiffusion: Controllable Disentangled Style Transfer via Diffusion Models15 Aug
  7. 2023IP-Adapter: Text Compatible Image Prompt Adapter for Text-to-Image Diffusion Models13 Aug
  8. 2023Designing a Better Asymmetric VQGAN for StableDiffusion7 Jun
  9. 2023Emergent Correspondence from Image Diffusion6 Jun
  10. 2023MultiDiffusion: Fusing Diffusion Paths for Controlled Image Generation16 Feb
  11. 2023Adding Conditional Control to Text-to-Image Diffusion Models10 Feb
  12. 2022Multi-Concept Customization of Text-to-Image Diffusion8 Dec
4 more
  1. 2022Inversion-Based Style Transfer with Diffusion Models23 Nov
  2. 2022DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation25 Aug
  3. 2022Blended Latent Diffusion6 Jun
  4. 2021More Control for Free! Image Synthesis with Semantic Diffusion Guidance10 Dec

Other references · 32

  1. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  2. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,380
  3. 2020Denoising Diffusion Implicit Modelscited by 9,781
  4. 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,255
  5. 2021Emerging Properties in Self-Supervised Vision Transformerscited by 6,968
  6. 2016Image-to-Image Translation with Conditional Adversarial Networkscited by 4,527
  7. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesiscited by 4,356
  8. 2016Perceptual Losses for Real-Time Style Transfer and Super-Resolutioncited by 2,694
  9. 2022An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversioncited by 2,245
  10. 2017Arbitrary Style Transfer in Real-time with Adaptive Instance Normalizationcited by 2,010
  11. 2021SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equationscited by 1,937
  12. 2022Plug-and-Play Diffusion Features for Text-Driven Image-to-Image Translationcited by 906
20 more
  1. 2023MasaCtrl: Tuning-Free Mutual Self-Attention Control for Consistent Image Synthesis and Editingcited by 727
  2. 2023Universal Guidance for Diffusion Modelscited by 437
  3. 2019SinGAN: Learning a Generative Model from a Single Natural Imagecited by 348
  4. 2023Diffusion Self-Guidance for Controllable Image Generationcited by 305
  5. 2023Style Aligned Image Generation via Shared Attentioncited by 216
  6. 2020Swapping Autoencoder for Deep Image Manipulationcited by 215
  7. 2022Splicing ViT Features for Semantic Appearance Transfercited by 209
  8. 2021StyTr2: Image Style Transfer with Transformerscited by 187
  9. 2024InstantStyle: Free Lunch towards Style-Preserving in Text-to-Image Generationcited by 170
  10. 2017Demystifying Neural Style Transfercited by 152
  11. 2023Cross-Image Attention for Zero-Shot Appearance Transfercited by 136
  12. 2018The Contextual Loss for Image Transformation with Non-Aligned Datacited by 135
  13. 2024Implicit Style-Content Separation using B-LoRAcited by 135
  14. 2016Combining Markov Random Fields and Convolutional Neural Networks for Image Synthesiscited by 102
  15. 2017Stable and Controllable Neural Texture Synthesis and Style Transfer Using Histogram Lossescited by 69
  16. 2018Non-Stationary Texture Synthesis by Adversarial Expansioncited by 67
  17. 2023StyleAdapter: A Unified Stylized Image Generation Modelcited by 39
  18. 2024Customizing Text-to-Image Models with a Single Image Paircited by 36
  19. 2022Generating natural images with direct Patch Distributions Matchingcited by 27
  20. 2024Generating Non-Stationary Textures using Self-Rectificationcited by 8

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 16; 2026: 17.

33 papers.

Sort:

By area:

  1. 2025REPA Works Until It Doesn't: Early-Stopped, Holistic Alignment Supercharges Diffusion Trainingcited by 37
  2. 2025Edit Transfer: Learning Image Editing via Vision In-Context Relationscited by 23
  3. 2025LookWhere? Efficient Visual Recognition by Learning Where to Look and What to See from Self-Supervisioncited by 5
  4. 2026MegaStyle: Constructing Diverse and Scalable Style Dataset via Consistent Text-to-Image Style Mappingcited by 5
  5. 2025Attention in Diffusion Model: A Surveycited by 4
  6. 2025Muon-Accelerated Attention Distillation for Real-Time Edge Synthesis via Optimized Latent Diffusioncited by 2
  7. 2026StyleGallery: Training-free and Semantic-aware Personalized Style Transfer from Arbitrary Image Referencescited by 2
  8. 2025Dynamic Importance in Diffusion U-Net for Enhanced Image Synthesiscited by 1
  9. 2025From Large to Super-Tiny: End-to-End Optimization for Cost-Efficient LLMscited by 1
  10. 2025OmniStyle2: Learning to Stylize by Learning to Destylizecited by 1
  11. 2026Mixture of Style Experts for Diverse Image Stylizationcited by 1
  12. 2026EditTransfer++: Toward Faithful and Efficient Visual-Prompt-Guided Image Editingcited by 1
21 more
  1. 2026Cut-ViT: Task-Specific Model Pruning via Gram Anchoring Subspace Consistencycited by 1
  2. 2025SSIMBaD: Sigma Scaling with SSIM-Guided Balanced Diffusion for AnimeFace Colorizationcited by 0
  3. 2025SPG: Style-Prompting Guidance for Style-Specific Content Creationcited by 0
  4. 2025Visually Grounded Narratives: Reducing Cognitive Burden in Researcher-Participant Interactioncited by 0
  5. 2025One-shot Embroidery Customization via Contrastive LoRA Modulationcited by 0
  6. 2025Image Generation Based on Image Style Extractioncited by 0
  7. 2025Text to Sketch Generation with Multi-Stylescited by 0
  8. 2025V-Shuffle: Zero-Shot Style Transfer via Value Shufflecited by 0
  9. 2025LumiX: Structured and Coherent Text-to-Intrinsic Generationcited by 0
  10. 2026DiffStyle3D: Consistent 3D Gaussian Stylization via Attention Optimizationcited by 0
  11. 2026DICE: Disentangling Artist Style from Content via Contrastive Subspace Decomposition in Diffusion Modelscited by 0
  12. 2026HAM: A Training-Free Style Transfer Approach via Heterogeneous Attention Modulation for Diffusion Modelscited by 0
  13. 2026Abstraction in Style: Beyond Texture and Colorcited by 0
  14. 2026Structured 3D Latents Are Surprisingly Powerful: Unleashing Generalizable Style with 2D Diffusioncited by 0
  15. 2026Semantic-Structural Alignment for Generative Pictorial Chartscited by 0
  16. 2026Capacity-Controlled Multi-View Stylization of 3D Gaussian Splattingcited by 0
  17. 2026PoseAlign: Sculpting Pose-Consistent Meshes via Text-Guided Deformationcited by 0
  18. 2026DReSG: Diffusion Residuals for Stylized Gaussian Splattingcited by 0
  19. 2026LoGAN: Multilingual Font Localization with Generative Agentscited by 0
  20. 2026CLeaR: A Unified Framework for Resolving the Leakage-Degradation Dilemma in Style Transfercited by 0
  21. 2026VAttnPO: Visual Attention-Guided Policy Optimizationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026TRACE: Time-Adaptive Residual Attention Control with Content-Style Decomposition for Training-Free Diffusion Style Transfer
  2. 2023Z^*: Zero-shot Style Transfer via Attention Rearrangement
  3. 2025DiCo: Revitalizing ConvNets for Scalable and Efficient Diffusion Modeling
  4. 2025Leveraging Diffusion Models for Stylization using Multiple Style Images
  5. 2026MAST: Mask-Guided Attention Control for Training-Free Regional-Multi Style Transfer
  6. 2024Diffusion Self-Distillation for Zero-Shot Customized Image Generation
  7. 2024Diff-2-in-1: Bridging Generation and Dense Perception with Diffusion Models
  8. 2024Enhancing Semantic Fidelity in Text-to-Image Synthesis: Attention Regulation in Diffusion Models