mlsanityData to 7 Oct 2026

Circumventing Concept Erasure Methods For Text-to-Image Generative Models

arXiv 2308.01508 ↗First posted 3 Aug 2023cs.LG cs.CR cs.CVcited by 81

Not on the map. Why?

Cited by 58 in Diffusion models, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

Text-to-image generative models can produce photo-realistic images for an extremely broad range of concepts, and their usage has proliferated widely among the general public. On the flip side, these models have numerous drawbacks, including their potential to generate images featuring sexually explicit content, mirror artistic styles without permission, or even hallucinate (or deepfake) the likenesses of celebrities. Consequently, various methods have been proposed in order to "erase" sensitive concepts from text-to-image models. In this work, we examine five recently proposed concept erasure methods, and show that targeted concepts are not fully excised from any of these methods. Specifically, we leverage the existence of special learned word embeddings that can retrieve "erased" concepts from the sanitized models with no alterations to their weights. Our results highlight the brittleness of post hoc concept erasure methods, and call into question their use in the algorithmic toolkit for AI safety.

Key facts. Cited by 81 papers in the map (data through 7 Oct 2026). Builds on 38 of its 49 references that are in the map. First posted on arXiv 3 Aug 2023.

Builds on

38 of 49 references.

References · 38

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  2. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  3. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
  4. 2020Denoising Diffusion Implicit Modelscited by 9,655
  5. 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,185
  6. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,005
  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. 2022Classifier-Free Diffusion Guidancecited by 5,717
  10. 2016Overcoming catastrophic forgetting in neural networkscited by 4,718
  11. 2021Zero-Shot Text-to-Image Generationcited by 4,276
  12. 2022LAION-5B: An open large-scale dataset for training next generation image-text modelscited by 4,131
26 more
  1. 2022DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generationcited by 3,286
  2. 2022An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversioncited by 2,231
  3. 2022Scaling Autoregressive Models for Content-Rich Text-to-Image Generationcited by 1,287
  4. 2017Continual Learning with Deep Generative Replaycited by 1,152
  5. 2022Mass-Editing Memory in a Transformercited by 976
  6. 2023Extracting Training Data from Diffusion Modelscited by 723
  7. 2023Muse: Text-To-Image Generation via Masked Generative Transformerscited by 640
  8. 2019Eternal Sunshine of the Spotless Net: Selective Forgetting in Deep Networkscited by 599
  9. 2023Erasing Concepts from Diffusion Modelscited by 530
  10. 2022Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Modelscited by 501
  11. 2022Make-A-Scene: Scene-Based Text-to-Image Generation with Human Priorscited by 487
  12. 2022Diffusion Art or Digital Forgery? Investigating Data Replication in Diffusion Modelscited by 418
  13. 2021GAN Inversion: A Surveycited by 366
  14. 2016Generative Visual Manipulation on the Natural Image Manifoldcited by 348
  15. 2023InstantBooth: Personalized Text-to-Image Generation without Test-Time Finetuningcited by 343
  16. 2023SVDiff: Compact Parameter Space for Diffusion Fine-Tuningcited by 338
  17. 2021Multimodal datasets: misogyny, pornography, and malignant stereotypescited by 326
  18. 2023Ablating Concepts in Text-to-Image Diffusion Modelscited by 312
  19. 2023Forget-Me-Not: Learning to Forget in Text-to-Image Diffusion Modelscited by 297
  20. 2022Red-Teaming the Stable Diffusion Safety Filtercited by 257
  21. 2023Understanding and Mitigating Copying in Diffusion Modelscited by 205
  22. 2022Versatile Diffusion: Text, Images and Variations All in One Diffusion Modelcited by 199
  23. 2023Selective Amnesia: A Continual Learning Approach to Forgetting in Deep Generative Modelscited by 175
  24. 2020Forgetting Outside the Box: Scrubbing Deep Networks of Information Accessible from Input-Output Observationscited by 171
  25. 2023Unsafe Diffusion: On the Generation of Unsafe Images and Hateful Memes From Text-To-Image Modelscited by 171
  26. 2023Editing Implicit Assumptions in Text-to-Image Diffusion Modelscited by 128

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 1; 2024: 23; 2025: 32; 2026: 25.

81 papers.

Sort:

By area:

  1. 2023To Generate or Not? Safety-Driven Unlearned Diffusion Models Are Still Easy To Generate Unsafe Images ... For Nowcited by 204
  2. 2024Defensive Unlearning with Adversarial Training for Robust Concept Erasure in Diffusion Modelscited by 164
  3. 2024SafeGen: Mitigating Sexually Explicit Content Generation in Text-to-Image Modelscited by 64
  4. 2024R.A.C.E.: Robust Adversarial Concept Erasure for Secure Text-to-Image Diffusion Modelcited by 56
  5. 2024ConceptPrune: Concept Editing in Diffusion Models via Skilled Neuron Pruningcited by 51
  6. 2024Direct Unlearning Optimization for Robust and Safe Text-to-Image Modelscited by 50
  7. 2024STEREO: A Two-Stage Framework for Adversarially Robust Concept Erasing from Text-to-Image Diffusion Modelscited by 39
  8. 2024Meta-Unlearning on Diffusion Models: Preventing Relearning Unlearned Conceptscited by 39
  9. 2025Fantastic Targets for Concept Erasure in Diffusion Models and Where To Find Themcited by 34
  10. 2024Dynamic Negative Guidance of Diffusion Modelscited by 28
  11. 2024Robust Concept Erasure Using Task Vectorscited by 27
  12. 2025Localized Concept Erasure for Text-to-Image Diffusion Models Using Training-Free Gated Low-Rank Adaptationcited by 27
38 more
  1. 2024TraSCE: Trajectory Steering for Concept Erasurecited by 25
  2. 2025CURE: Concept Unlearning via Orthogonal Representation Editing in Diffusion Modelscited by 24
  3. 2024Holistic Unlearning Benchmark: A Multi-Faceted Evaluation for Text-to-Image Diffusion Model Unlearningcited by 19
  4. 2025Training-Free Safe Denoisers for Safe Use of Diffusion Modelscited by 19
  5. 2024FreeCompose: Generic Zero-Shot Image Composition with Diffusion Priorcited by 16
  6. 2024Espresso: Robust Concept Filtering in Text-to-Image Modelscited by 14
  7. 2025One Image is Worth a Thousand Words: A Usability Preservable Text-Image Collaborative Erasing Frameworkcited by 14
  8. 2025When Are Concepts Erased From Diffusion Models?cited by 14
  9. 2024Probing Unlearned Diffusion Models: A Transferable Adversarial Attack Perspectivecited by 13
  10. 2025Yuan: Yielding Unblemished Aesthetics Through A Unified Network for Visual Imperfections Removal in Generated Imagescited by 13
  11. 2024Unlearning or Concealment? A Critical Analysis and Evaluation Metrics for Unlearning in Diffusion Modelscited by 12
  12. 2025Erasing More Than Intended? How Concept Erasure Degrades the Generation of Non-Target Conceptscited by 12
  13. 2025Detect-and-Guide: Self-regulation of Diffusion Models for Safe Text-to-Image Generation via Guideline Token Optimizationcited by 12
  14. 2024Replication in Visual Diffusion Models: A Survey and Outlookcited by 11
  15. 2024Unveiling Concept Attribution in Diffusion Modelscited by 11
  16. 2025Erasing Concepts, Steering Generations: A Comprehensive Survey of Concept Suppressioncited by 10
  17. 2024Memories of Forgotten Conceptscited by 9
  18. 2025A Comprehensive Survey on Concept Erasure in Text-to-Image Diffusion Modelscited by 9
  19. 2025Erased but Not Forgotten: How Backdoors Compromise Concept Erasurecited by 9
  20. 2025Erased or Dormant? Rethinking Concept Erasure Through Reversibilitycited by 9
  21. 2025Rethinking Machine Unlearning in Image Generation Modelscited by 9
  22. 2025Image Can Bring Your Memory Back: A Novel Multi-Modal Guided Attack against Image Generation Model Unlearningcited by 9
  23. 2024A Survey on Responsible Generative AI: What to Generate and What Notcited by 8
  24. 2025SuMa: A Subspace Mapping Approach for Robust and Effective Concept Erasure in Text-to-Image Diffusion Modelscited by 8
  25. 2025EraseFlow: Learning Concept Erasure Policies via GFlowNet-Driven Alignmentcited by 8
  26. 2026TraceRouter: Robust Safety for Large Foundation Models via Path-Level Interventioncited by 8
  27. 2024Model Integrity when Unlearning with T2I Diffusion Modelscited by 7
  28. 2025CE-SDWV: Effective and Efficient Concept Erasure for Text-to-Image Diffusion Models via a Semantic-Driven Word Vocabularycited by 7
  29. 2025Mitigating Sexual Content Generation via Embedding Distortion in Text-conditioned Diffusion Modelscited by 7
  30. 2025Training-Free Safe Text Embedding Guidance for Text-to-Image Diffusion Modelscited by 6
  31. 2024Safe Text-to-Image Generation: Simply Sanitize the Prompt Embeddingcited by 4
  32. 2025EMMA: Concept Erasure Benchmark with Comprehensive Semantic Metrics and Diverse Categoriescited by 4
  33. 2026Co-occurring Associated REtained concepts in Diffusion Unlearningcited by 4
  34. 2025Now You See It, Now You Don't - Instant Concept Erasure for Safe Text-to-Image and Video Generationcited by 3
  35. 2026AEGIS: Adversarial Target-Guided Retention-Data-Free Robust Concept Erasure from Diffusion Modelscited by 3
  36. 2026OrthoEraser: Coupled-Neuron Orthogonal Projection for Concept Erasurecited by 3
  37. 2026To Erase, or Not to Erase: Robust Training-Free Concept Erasure with Preservation aware Adaptive Ranked Subspace Expansioncited by 3
  38. 2025Rethinking the Vulnerability of Concept Erasure and a New Methodcited by 2

50 most cited of 81 All 81 (JSON)