mlsanityData to 7 Oct 2026

TIDE : Temporal-Aware Sparse Autoencoders for Interpretable Diffusion Transformers in Image Generation

arXiv 2503.07050 ↗First posted 10 Mar 2025cs.CV cs.AI cs.MMcited by 16

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

Diffusion Transformers (DiTs) are a powerful yet underexplored class of generative models compared to U-Net-based diffusion architectures. We propose TIDE-Temporal-aware sparse autoencoders for Interpretable Diffusion transformErs-a framework designed to extract sparse, interpretable activation features across timesteps in DiTs. TIDE effectively captures temporally-varying representations and reveals that DiTs naturally learn hierarchical semantics (e.g., 3D structure, object class, and fine-grained concepts) during large-scale pretraining. Experiments show that TIDE enhances interpretability and controllability while maintaining reasonable generation quality, enabling applications such as safe image editing and style transfer.

Key facts. Cited by 16 papers in the map (data through 7 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 44 of its 68 references that are in the map. First posted on arXiv 10 Mar 2025.

Builds on

44 of 68 references.

Closest ancestors · 15

Same direction, newest first

  1. 2024textitRevelio: Interpreting and leveraging semantic information in diffusion models23 Nov
  2. 2024One-Step is Enough: Sparse Autoencoders for Text-to-Image Diffusion Models28 Oct
  3. 2024SANA: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformers14 Oct
  4. 2024Isometric Representation Learning for Disentangled Latent Space of Diffusion Models16 Jul
  5. 2024Lumina-Next: Making Lumina-T2X Stronger and Faster with Next-DiT5 Jun
  6. 2024Scaling Rectified Flow Transformers for High-Resolution Image Synthesis5 Mar
  7. 2023PixArt-α: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis30 Sep
  8. 2023Understanding the Latent Space of Diffusion Models through the Lens of Riemannian Geometry24 Jul
  9. 2023Diffusion Models Beat GANs on Image Classification17 Jul
  10. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis4 Jul
  11. 2022Scalable Diffusion Models with Transformers19 Dec
  12. 2022Diffusion Models already have a Semantic Latent Space20 Oct
3 more
  1. 2022DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps2 Jun
  2. 2021High-Resolution Image Synthesis with Latent Diffusion Models20 Dec
  3. 2020Denoising Diffusion Implicit Models6 Oct

Other references · 29

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  2. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,604
  3. 2014Microsoft COCO: Common Objects in Contextcited by 17,498
  4. 2014Going Deeper with Convolutionscited by 4,231
  5. 2019Generative Modeling by Estimating Gradients of the Data Distributioncited by 4,067
  6. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,359
  7. 2024Scaling and evaluating sparse autoencoderscited by 530
  8. 2022Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Modelscited by 501
  9. 2024Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Modelscited by 412
  10. 2023Unified Concept Editing in Diffusion Modelscited by 377
  11. 2023Repurposing Diffusion-Based Image Generators for Monocular Depth Estimationcited by 323
  12. 2023Your Diffusion Model is Secretly a Zero-Shot Classifiercited by 301
17 more
  1. 2022What the DAAM: Interpreting Stable Diffusion Using Cross Attentioncited by 233
  2. 2013k-Sparse Autoencoderscited by 209
  3. 2021EditGAN: High-Precision Semantic Image Editingcited by 156
  4. 2023Diffuse, Attend, and Segment: Unsupervised Zero-Shot Segmentation using Stable Diffusioncited by 131
  5. 2024SAFREE: Training-Free and Adaptive Guard for Safe Text-to-Image And Video Generationcited by 125
  6. 2021Transformer visualization via dictionary learning: contextualized embedding as a linear superposition of transformer factorscited by 106
  7. 2024Interpreting and Editing Vision-Language Representations to Mitigate Hallucinationscited by 103
  8. 2024Discover-then-Name: Task-Agnostic Concept Bottlenecks via Automated Concept Discoverycited by 91
  9. 2024DepthFM: Fast Monocular Depth Estimation with Flow Matchingcited by 76
  10. 2023Localizing and Editing Knowledge in Text-to-Image Generative Modelscited by 52
  11. 2024Towards Understanding the Working Mechanism of Text-to-Image Diffusion Modelcited by 48
  12. 2024The Missing Curve Detectors of InceptionV1: Applying Sparse Autoencoders to InceptionV1 Early Visioncited by 37
  13. 2024Diffusion Lens: Interpreting Text Encoders in Text-to-Image Pipelinescited by 28
  14. 2024On Mechanistic Knowledge Localization in Text-to-Image Generative Modelscited by 27
  15. 2024STIV: Scalable Text and Image Conditioned Video Generationcited by 24
  16. 2024An X-Ray Is Worth 15 Features: Sparse Autoencoders for Interpretable Radiology Report Generationcited by 23
  17. 2025Lumina-Video: Efficient and Flexible Video Generation with Multi-scale Next-DiTcited by 16

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 5; 2026: 11.

16 papers.

Sort:

By area:

  1. 2025JPS: Jailbreak Multimodal Large Language Models with Collaborative Visual Perturbation and Textual Steeringcited by 10
  2. 2025Temporal Concept Dynamics in Diffusion Models via Prompt-Conditioned Interventionscited by 7
  3. 2025Vision-to-Music Generation: A Surveycited by 4
  4. 2026CASL: Concept-Aligned Sparse Latents for Interpreting Diffusion Modelscited by 2
  5. 2026ELROND: Exploring and decomposing intrinsic capabilities of diffusion modelscited by 2
  6. 2026DifFRACT: Diffusion Feature Reconstruction and Attribution for Circuit Tracingcited by 2
  7. 2025LouvreSAE: Sparse Autoencoders for Interpretable and Controllable Style Transfercited by 1
  8. 2026Driving Intents Amplify Planning-Oriented Reinforcement Learningcited by 1
  9. 2026Residualized Temporal Sparse Autoencoders for Interpreting Diffusion Modelscited by 1
  10. 2025Beyond the Black Box: Identifiable Interpretation and Control in Generative Models via Causal Minimalitycited by 0
  11. 2026Latent Anomaly Knowledge Excavation: Unveiling Sparse Sensitive Neurons in Vision-Language Modelscited by 0
  12. 2026SMolLM: Small Language Models Learn Small Molecular Grammarcited by 0
4 more
  1. 2026Robust and Generalizable Safety Steering for Text-to-Image Diffusion Transformerscited by 0
  2. 2026V-FIND: Revealing the Intrinsic Forgery Knowledge Encoded in Video Forgery Detectorscited by 0
  3. 2026Why Do Video Diffusion Models Violate Physics? Unveiling the Flaws in Attention Mechanismscited by 0
  4. 2026D-Scope: Decomposing and Steering Diffusion Transformers with Sparse Autoencoderscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Diffusion Transformers with Representation Autoencoders
  2. 2025Emergence and Evolution of Interpretable Concepts in Diffusion Models
  3. 2026Look But Don't Touch with Sparse Autoencoders for Unlearning in Diffusion Models
  4. 2025Boosting Generative Image Modeling via Joint Image-Feature Synthesis
  5. 2026MMDiff: Extending Diffusion Transformers for Multi-Modal Generation
  6. 2023Hierarchical Diffusion Autoencoders and Disentangled Image Manipulation
  7. 2026A Hidden Semantic Bottleneck in Conditional Embeddings of Diffusion Transformers
  8. 2025Interpreting Large Text-to-Image Diffusion Models with Dictionary Learning