mlsanityData to 14 Oct 2026

EraseAnything: Enabling Concept Erasure in Rectified Flow Transformers

arXiv 2412.20413 ↗First posted 29 Dec 2024cs.CVcited by 95

On the map: Diffusion models → Robustness, privacy and adversarial aspects

Abstract (arXiv)

Removing unwanted concepts from large-scale text-to-image (T2I) diffusion models while maintaining their overall generative quality remains an open challenge. This difficulty is especially pronounced in emerging paradigms, such as Stable Diffusion (SD) v3 and Flux, which incorporate flow matching and transformer-based architectures. These advancements limit the transferability of existing concept-erasure techniques that were originally designed for the previous T2I paradigm (e.g., SD v1.4). In this work, we introduce EraseAnything, the first method specifically developed to address concept erasure within the latest flow-based T2I framework. We formulate concept erasure as a bi-level optimization problem, employing LoRA-based parameter tuning and an attention map regularizer to selectively suppress undesirable activations. Furthermore, we propose a self-contrastive learning strategy to ensure that removing unwanted concepts does not inadvertently harm performance on unrelated ones. Experimental results demonstrate that EraseAnything successfully fills the research gap left by earlier methods in this new T2I paradigm, achieving state-of-the-art performance across a wide range of concept erasure tasks.

Key facts. Cited by 95 papers in the map (data through 14 Oct 2026). Placed in Robustness, privacy and adversarial aspects, one of 14 directions in Diffusion models. Builds on 42 of its 51 references that are in the map. First posted on arXiv 29 Dec 2024.

Builds on

42 of 51 references.

Closest ancestors · 11

Same direction, newest first

  1. 2024Erasing Undesirable Concepts in Diffusion Models with Adversarial Preservation21 Oct
  2. 2024Meta-Unlearning on Diffusion Models: Preventing Relearning Unlearned Concepts16 Oct
  3. 2024Defensive Unlearning with Adversarial Training for Robust Concept Erasure in Diffusion Models24 May
  4. 2024MACE: Mass Concept Erasure in Diffusion Models10 Mar
  5. 2023One-Dimensional Adapter to Rule Them All: Concepts, Diffusion Models and Erasing Applications26 Dec
  6. 2023Receler: Reliable Concept Erasing of Text-to-Image Diffusion Models via Lightweight Erasers29 Nov
  7. 2023Unified Concept Editing in Diffusion Models25 Aug
  8. 2023Ablating Concepts in Text-to-Image Diffusion Models23 Mar
  9. 2023Erasing Concepts from Diffusion Models13 Mar
  10. 2022Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Models9 Nov
  11. 2022Red-Teaming the Stable Diffusion Safety Filter3 Oct

Other references · 31

  1. 2017Attention Is All You Needcited by 60,361
  2. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  3. 2020Denoising Diffusion Probabilistic Modelscited by 21,076
  4. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,713
  5. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  6. 2023GPT-4 Technical Reportcited by 18,396
  7. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  8. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,599
  9. 2020Denoising Diffusion Implicit Modelscited by 9,781
  10. 2018Representation Learning with Contrastive Predictive Codingcited by 7,777
  11. 2019Momentum Contrast for Unsupervised Visual Representation Learningcited by 7,513
  12. 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,585
19 more
  1. 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 6,355
  2. 2022Flow Matching for Generative Modelingcited by 5,702
  3. 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,703
  4. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesiscited by 4,356
  5. 2021Zero-Shot Text-to-Image Generationcited by 4,288
  6. 2024Scaling Rectified Flow Transformers for High-Resolution Image Synthesiscited by 4,181
  7. 2022LAION-5B: An open large-scale dataset for training next generation image-text modelscited by 4,152
  8. 2022Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flowcited by 3,700
  9. 2021GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Modelscited by 3,614
  10. 2022Prompt-to-Prompt Image Editing with Cross Attention Controlcited by 2,431
  11. 2017Decoupled Weight Decay Regularizationcited by 1,831
  12. 2018Bilevel Programming for Hyperparameter Optimization and Meta-Learningcited by 405
  13. 2019Optimizing Millions of Hyperparameters by Implicit Differentiationcited by 304
  14. 2023BoxDiff: Text-to-Image Synthesis with Training-Free Box-Constrained Diffusioncited by 282
  15. 2017A Review on Bilevel Optimization: From Classical to Evolutionary Approaches and Applicationscited by 129
  16. 2023ReVersion: Diffusion-Based Relation Inversion from Imagescited by 82
  17. 2022Bi-level Physics-Informed Neural Networks for PDE Constrained Optimization using Broyden's Hypergradientscited by 17
  18. 2024RealEra: Semantic-level Concept Erasure via Neighbor-Concept Miningcited by 15
  19. 2024Memory-Efficient Gradient Unrolling for Large-Scale Bi-level Optimizationcited by 11

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 48; 2026: 47.

95 papers.

Sort:

By area:

  1. 2025In-Context Edit: Enabling Instructional Image Editing with In-Context Generation in Large Scale Diffusion Transformercited by 204
  2. 2025Set You Straight: Auto-Steering Denoising Trajectories to Sidestep Unwanted Conceptscited by 62
  3. 2025GestureLSM: Latent Shortcut based Co-Speech Gesture Generation with Spatial-Temporal Modelingcited by 44
  4. 2025OFFSET: Segmentation-based Focus Shift Revision for Composed Image Retrievalcited by 40
  5. 2025Does FLUX Already Know How to Perform Physically Plausible Image Composition?cited by 32
  6. 2025HUD: Hierarchical Uncertainty-Aware Disambiguation Network for Composed Video Retrievalcited by 32
  7. 2025Towards Universal Modal Tracking with Online Dense Temporal Token Learningcited by 30
  8. 2025DragFlow: Unleashing DiT Priors with Region Based Supervision for Drag Editingcited by 29
  9. 2025DreamRenderer: Taming Multi-Instance Attribute Control in Large-Scale Text-to-Image Modelscited by 28
  10. 2025UltraHR-100K: Enhancing UHR Image Synthesis with A Large-Scale High-Quality Datasetcited by 18
  11. 2025One Image is Worth a Thousand Words: A Usability Preservable Text-Image Collaborative Erasing Frameworkcited by 16
  12. 2026LUVE : Latent-Cascaded Ultra-High-Resolution Video Generation with Dual Frequency Expertscited by 15
38 more
  1. 2025Erasing More Than Intended? How Concept Erasure Degrades the Generation of Non-Target Conceptscited by 12
  2. 2025Contextual Gesture: Co-Speech Gesture Video Generation through Context-aware Gesture Representationcited by 12
  3. 2025CRISP-SAM2: SAM2 with Cross-Modal Interaction and Semantic Prompting for Multi-Organ Segmentationcited by 12
  4. 2025Show and Polish: Reference-Guided Identity Preservation in Face Video Restorationcited by 12
  5. 2025Temporal-Guided Spiking Neural Networks for Event-Based Human Action Recognitioncited by 11
  6. 2025Revoking Amnesia: RL-based Trajectory Optimization to Resurrect Erased Concepts in Diffusion Modelscited by 11
  7. 2025Erasing Concepts, Steering Generations: A Comprehensive Survey of Concept Suppressioncited by 10
  8. 2025SAEmnesia: Erasing Concepts in Diffusion Models with Supervised Sparse Autoencoderscited by 10
  9. 2025Erased or Dormant? Rethinking Concept Erasure Through Reversibilitycited by 9
  10. 2025EraseFlow: Learning Concept Erasure Policies via GFlowNet-Driven Alignmentcited by 9
  11. 2025CE-SDWV: Effective and Efficient Concept Erasure for Text-to-Image Diffusion Models via a Semantic-Driven Word Vocabularycited by 8
  12. 2025Intentional Gesture: Deliver Your Intentions with Gestures for Speechcited by 8
  13. 2025Mitigating Sexual Content Generation via Embedding Distortion in Text-conditioned Diffusion Modelscited by 7
  14. 2026From Zero to Detail: A Progressive Spectral Decoupling Paradigm for UHD Image Restoration with New Benchmarkcited by 7
  15. 2025Subject Information Extraction for Novelty Detection with Domain Shiftscited by 6
  16. 2026Differential Vector Erasure: Unified Training-Free Concept Erasure for Flow Matching Modelscited by 5
  17. 2025Enhancing Zero-Shot Brain Tumor Subtype Classification via Fine-Grained Patch-Text Alignmentcited by 4
  18. 2025Lego-Edit: A General Image Editing Framework with Model-Level Bricks and MLLM Buildercited by 4
  19. 2025Diffusion-Based Image Editing for Breaking Robust Watermarkscited by 4
  20. 2026ActErase: A Training-Free Paradigm for Precise Concept Erasure via Activation Redirectioncited by 4
  21. 2025ACE: Attentional Concept Erasure in Diffusion Modelscited by 3
  22. 2025Erased, But Not Forgotten: Erased Rectified Flow Transformers Still Remain Unsafe Under Concept Attackcited by 3
  23. 2025Diffusion-Based Image Editing: An Unforeseen Adversary to Robust Invisible Watermarkscited by 3
  24. 2025CGCE: Classifier-Guided Concept Erasure in Generative Modelscited by 3
  25. 2026UnHype: CLIP-Guided Hypernetworks for Dynamic LoRA Unlearningcited by 3
  26. 2026Localized Concept Erasure in Text-to-Image Diffusion Models via High-Level Representation Misdirectioncited by 3
  27. 2026Closed-Form Concept Erasure via Double Projectionscited by 3
  28. 2025Rethinking the Vulnerability of Concept Erasure and a New Methodcited by 2
  29. 2025FLUXSynID: A Framework for Identity-Controlled Synthetic Face Generation with Document and Live Imagescited by 2
  30. 2025TRACE: Trajectory-Constrained Concept Erasure in Diffusion Modelscited by 2
  31. 2025VCE: Safe Autoregressive Image Generation via Visual Contrast Exploitationcited by 2
  32. 2025SSCM: A Spatial-Semantic Consistent Model for Multi-Contrast MRI Super-Resolutioncited by 2
  33. 2025GrOCE:Graph-Guided Online Concept Erasure for Text-to-Image Diffusion Modelscited by 2
  34. 2026Vanishing Watermarks: Diffusion-Based Image Editing Undermines Robust Invisible Watermarkingcited by 2
  35. 2026SafeRoPE: Risk-specific Head-wise Embedding Rotation for Safe Generation in Rectified Flow Transformerscited by 2
  36. 2026FlowErase-RL: Rethinking Concept Erasure as Reward Optimization in Flow Matching Modelscited by 2
  37. 2026Orthogonal Concept Erasure for Diffusion Modelscited by 2
  38. 2026GEM: Geometric Erasure by Contrastive Velocity Matching in Rectified Flowscited by 2

50 most cited of 95 All 95 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026Z-Erase: Enabling Concept Erasure in Single-Stream Diffusion Transformers
  2. 2025SPEED: Scalable, Precise, and Efficient Concept Erasure for Diffusion Models
  3. 2026Erasing Without Collateral Damage: Precise Concept Removal in Diffusion Models
  4. 2026Erasing Thousands of Concepts: Towards Scalable and Practical Concept Erasure for Text-to-Image Diffusion Models
  5. 2024Precise, Fast, and Low-cost Concept Erasure in Value Space: Orthogonal Complement Matters
  6. 2025Concept Corrector: Erase concepts on the fly for text-to-image diffusion models
  7. 2026Prototype-Guided Concept Erasure in Diffusion Models
  8. 2025Semantic Surgery: Zero-Shot Concept Erasure in Diffusion Models