mlsanityData to 7 Oct 2026

textitRevelio: Interpreting and leveraging semantic information in diffusion models

arXiv 2411.16725 ↗First posted 23 Nov 2024cs.CVcited by 29

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

We study textithow rich visual semantic information is represented within various layers and denoising timesteps of different diffusion architectures. We uncover monosemantic interpretable features by leveraging k-sparse autoencoders (k-SAE). We substantiate our mechanistic interpretations via transfer learning using light-weight classifiers on off-the-shelf diffusion models' features. On 4 datasets, we demonstrate the effectiveness of diffusion features for representation learning. We provide an in-depth analysis of how different diffusion architectures, pre-training datasets, and language model conditioning impacts visual representation granularity, inductive biases, and transfer learning capabilities. Our work is a critical step towards deepening interpretability of black-box diffusion models. Code and visualizations available at: https://github.com/revelio-diffusion/revelio

Key facts. Cited by 29 papers in the map (data through 7 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 46 of its 70 references that are in the map. First posted on arXiv 23 Nov 2024.

Builds on

46 of 70 references.

Closest ancestors · 9

Same direction, newest first

  1. 2023Diffusion Model as Representation Learner21 Aug
  2. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis4 Jul
  3. 2022Scalable Diffusion Models with Transformers19 Dec
  4. 2022Diffusion Models already have a Semantic Latent Space20 Oct
  5. 2022All are Worth Words: A ViT Backbone for Diffusion Models25 Sep
  6. 2021High-Resolution Image Synthesis with Latent Diffusion Models20 Dec
  7. 2021Diffusion Models Beat GANs on Image Synthesis11 May
  8. 2020Denoising Diffusion Implicit Models6 Oct
  9. 2020Denoising Diffusion Probabilistic Models19 Jun

Other references · 37

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  3. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  4. 2023GPT-4 Technical Reportcited by 18,132
  5. 2023Visual Instruction Tuningcited by 7,792
  6. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,764
  7. 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,539
  8. 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 6,322
  9. 2013Fine-Grained Visual Classification of Aircraftcited by 1,701
  10. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,359
  11. 2023MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilitiescited by 1,221
  12. 2022Reproducible scaling laws for contrastive language-image learningcited by 1,163
25 more
  1. 2022Plug-and-Play Diffusion Features for Text-Driven Image-to-Image Translationcited by 897
  2. 2024Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMscited by 719
  3. 2024Scaling and evaluating sparse autoencoderscited by 530
  4. 2021Label-Efficient Semantic Segmentation with Diffusion Modelscited by 495
  5. 2023Open-Vocabulary Panoptic Segmentation with Text-to-Image Diffusion Modelscited by 371
  6. 2022DiffusionDet: Diffusion Model for Object Detectioncited by 354
  7. 2023Your Diffusion Model is Secretly a Zero-Shot Classifiercited by 301
  8. 2023Unleashing Text-to-Image Diffusion Models for Visual Perceptioncited by 266
  9. 2023Emu Video: Factorizing Text-to-Video Generation by Explicit Image Conditioningcited by 254
  10. 2023P+: Extended Textual Conditioning in Text-to-Image Generationcited by 242
  11. 2023Diffusion Hyperfeatures: Searching Through Time and Space for Semantic Correspondencecited by 211
  12. 2013k-Sparse Autoencoderscited by 209
  13. 2023Localizing Object-level Shape Variations with Text-to-Image Diffusion Modelscited by 142
  14. 2023Text-to-Image Diffusion Models are Zero-Shot Classifierscited by 140
  15. 2023DatasetDM: Synthesizing Data with Perception Annotations Using Diffusion Modelscited by 122
  16. 2021Generative Models as a Data Source for Multiview Representation Learningcited by 104
  17. 2024Understanding the Impact of Negative Prompts: When and How Do They Take Effect?cited by 50
  18. 2023FeatureNeRF: Learning Generalizable NeRFs by Distilling Foundation Modelscited by 47
  19. 2024Diffusion Feedback Helps CLIP See Bettercited by 37
  20. 2017Weakly-supervised learning of visual relationscited by 36
  21. 2023The Hidden Language of Diffusion Modelscited by 34
  22. 2024Diffusion Lens: Interpreting Text Encoders in Text-to-Image Pipelinescited by 28
  23. 2024On Mechanistic Knowledge Localization in Text-to-Image Generative Modelscited by 27
  24. 2024DiffusionPID: Interpreting Diffusion via Partial Information Decompositioncited by 16
  25. 2023Discffusion: Discriminative Diffusion Models as Few-shot Vision and Language Learnerscited by 14

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 13; 2026: 16.

29 papers.

Sort:

By area:

  1. 2025SAeUron: Interpretable Concept Unlearning in Diffusion Models with Sparse Autoencoderscited by 75
  2. 2025Sparse Autoencoders Learn Monosemantic Features in Vision-Language Modelscited by 70
  3. 2025TIDE : Temporal-Aware Sparse Autoencoders for Interpretable Diffusion Transformers in Image Generationcited by 16
  4. 2025SAEmnesia: Erasing Concepts in Diffusion Models with Supervised Sparse Autoencoderscited by 8
  5. 2026AudioSAE: Towards Understanding of Audio-Processing Models with Sparse AutoEncoderscited by 8
  6. 2025LayerSync: Self-aligning Intermediate Layerscited by 7
  7. 2025Concept Steerers: Leveraging K-Sparse Autoencoders for Test-Time Controllable Generationscited by 6
  8. 2025Interpreting Large Text-to-Image Diffusion Models with Dictionary Learningcited by 6
  9. 2025SUB: Benchmarking CBM Generalization via Synthetic Attribute Substitutionscited by 6
  10. 2025Interpretable and Steerable Concept Bottleneck Sparse Autoencoderscited by 6
  11. 2025ConceptScope: Characterizing Dataset Bias via Disentangled Visual Conceptscited by 5
  12. 2025What's in a Latent? Leveraging Diffusion Latent Space for Domain Generalizationcited by 3
17 more
  1. 2026DDiT: Dynamic Patch Scheduling for Efficient Diffusion Transformerscited by 3
  2. 2026Towards Interpretable Hallucination Analysis and Mitigation in LVLMs via Contrastive Neuron Steeringcited by 2
  3. 2026ELROND: Exploring and decomposing intrinsic capabilities of diffusion modelscited by 2
  4. 2026DeVA: Decoupled Video-Action Model with physical guidance for robot policy learningcited by 2
  5. 2025Characterizing Motion Encoding in Video Diffusion Timestepscited by 1
  6. 2026Diffusion Representations for Fine-Grained Image Classification: A Marine Plankton Case Studycited by 1
  7. 2026Guiding a Diffusion Model by Swapping Its Tokenscited by 1
  8. 2025Beyond the Black Box: Identifiable Interpretation and Control in Generative Models via Causal Minimalitycited by 0
  9. 2026Latent Structure Emergence in Diffusion Models via Confidence-Based Filteringcited by 0
  10. 2026RAIGen: Rare Attribute Identification in Text-to-Image Generative Modelscited by 0
  11. 2026The Latent Color Subspace: Emergent Order in High-Dimensional Chaoscited by 0
  12. 2026Revisiting Autoregressive Models for Generative Image Classificationcited by 0
  13. 2026Information-Theoretic Classifier-Free Guidance with Adaptive Schedule Optimizationcited by 0
  14. 2026Attention Dynamics in Diffusion Models: A Visual Analytics Framework for Human-AI Collaborationcited by 0
  15. 2026SCOUT: Semantic Concept Discovery for Open-Vocabulary Editing of face Recognition Templatescited by 0
  16. 2026RolloutFaith: Auditing Persistent Internal Interventions in Visual World Modelcited by 0
  17. 2026DiffReID: Discriminative Diffusion Model for Object Re-Identificationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Emergence and Evolution of Interpretable Concepts in Diffusion Models
  2. 2026Look But Don't Touch with Sparse Autoencoders for Unlearning in Diffusion Models
  3. 2024Decoding Diffusion: A Scalable Framework for Unsupervised Analysis of Latent Space Biases and Representations Using Natural Language Prompts
  4. 2025Patronus: Interpretable Diffusion Models with Prototypes
  5. 2025Automated Learning of Semantic Embedding Representations for Diffusion Models
  6. 2026Deep Dreams Are Made of This: Visualizing Monosemantic Features in Diffusion Models
  7. 2025Generalization of Diffusion Models Arises with a Balanced Representation Space
  8. 2026DLM-Scope: Mechanistic Interpretability of Diffusion Language Models via Sparse Autoencoders