mlsanityData to 7 Oct 2026

AEGIS: Adversarial Target-Guided Retention-Data-Free Robust Concept Erasure from Diffusion Models

arXiv 2602.06771 ↗First posted 6 Feb 2026cs.LG cs.AI cs.CRcited by 3

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

also in 1 more
Abstract (arXiv)

Concept erasure helps stop diffusion models (DMs) from generating harmful content; but current methods face robustness retention trade off. Robustness means the model fine-tuned by concept erasure methods resists reactivation of erased concepts, even under semantically related prompts. Retention means unrelated concepts are preserved so the model's overall utility stays intact. Both are critical for concept erasure in practice, yet addressing them simultaneously is challenging, as existing works typically improve one factor while sacrificing the other. Prior work typically strengthens one while degrading the other, e.g., mapping a single erased prompt to a fixed safe target leaves class level remnants exploitable by prompt attacks, whereas retention-oriented schemes underperform against adaptive adversaries. This paper introduces Adversarial Erasure with Gradient Informed Synergy (AEGIS), a retention-data-free framework that advances both robustness and retention.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Placed in Robustness, privacy and adversarial aspects, one of 14 directions in Diffusion models. Builds on 51 of its 57 references that are in the map. First posted on arXiv 6 Feb 2026.

Builds on

51 of 57 references.

Closest ancestors · 19

Same direction, newest first

  1. 2025Fantastic Targets for Concept Erasure in Diffusion Models and Where To Find Them31 Jan
  2. 2024Erasing Undesirable Concepts in Diffusion Models with Adversarial Preservation21 Oct
  3. 2024STEREO: A Two-Stage Framework for Adversarially Robust Concept Erasing from Text-to-Image Diffusion Models29 Aug
  4. 2024Reliable and Efficient Concept Erasure of Text-to-Image Diffusion Models17 Jul
  5. 2024Defensive Unlearning with Adversarial Training for Robust Concept Erasure in Diffusion Models24 May
  6. 2024Probing Unlearned Diffusion Models: A Transferable Adversarial Attack Perspective30 Apr
  7. 2024MACE: Mass Concept Erasure in Diffusion Models10 Mar
  8. 2023One-Dimensional Adapter to Rule Them All: Concepts, Diffusion Models and Erasing Applications26 Dec
  9. 2023SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation19 Oct
  10. 2023To Generate or Not? Safety-Driven Unlearned Diffusion Models Are Still Easy To Generate Unsafe Images ... For Now18 Oct
  11. 2023Ring-A-Bell! How Reliable are Concept Removal Methods for Diffusion Models?16 Oct
  12. 2023Prompting4Debugging: Red-Teaming Text-to-Image Diffusion Models by Finding Problematic Prompts12 Sep
7 more
  1. 2023Unified Concept Editing in Diffusion Models25 Aug
  2. 2023Robust Classification via a Single Diffusion Model24 May
  3. 2023SneakyPrompt: Jailbreaking Text-to-image Generative Models20 May
  4. 2023Forget-Me-Not: Learning to Forget in Text-to-Image Diffusion Models30 Mar
  5. 2023Ablating Concepts in Text-to-Image Diffusion Models23 Mar
  6. 2023Erasing Concepts from Diffusion Models13 Mar
  7. 2022Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Models9 Nov

Other references · 32

  1. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  2. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
  3. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,005
  4. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
  5. 2022LAION-5B: An open large-scale dataset for training next generation image-text modelscited by 4,131
  6. 2020Gradient Surgery for Multi-Task Learningcited by 1,210
  7. 2023Hard Prompts Made Easy: Gradient-Based Discrete Optimization for Prompt Tuning and Discoverycited by 329
  8. 2023Editing Implicit Assumptions in Text-to-Image Diffusion Modelscited by 128
  9. 2020Improving Image Captioning with Better Use of Captionscited by 84
  10. 2023Circumventing Concept Erasure Methods For Text-to-Image Generative Modelscited by 81
  11. 2024Get What You Want, Not What You Don't: Image Content Suppression for Text-to-Image Diffusion Modelscited by 62
  12. 2024Scissorhands: Scrub Data Influence via Connection Sensitivity in Networkscited by 50
20 more
  1. 2022Towards Robust Rain Removal Against Adversarial Attacks: A Comprehensive Benchmark Analysis and Beyondcited by 48
  2. 2021Boosting Fast Adversarial Training with Learnable Adversarial Initializationcited by 40
  3. 2023Backdoor Attacks Against Deep Image Compression via Adaptive Frequency Triggercited by 40
  4. 2022Prior-Guided Adversarial Initialization for Fast Adversarial Trainingcited by 34
  5. 2022Fast Adversarial Training with Adaptive Step Sizecited by 24
  6. 2024Transferable Adversarial Attacks on SAM and Its Downstream Modelscited by 20
  7. 2024Purify Unlearnable Examples via Rate-Constrained Variational Autoencoderscited by 18
  8. 2025LLM Unlearning with LLM Beliefscited by 17
  9. 2024Mitigating the Curse of Dimensionality for Certified Robustness via Dual Randomized Smoothingcited by 15
  10. 2024Semantic Deep Hiding for Robust Unlearnable Examplescited by 13
  11. 2024DAT: Improving Adversarial Robustness via Generative Amplitude Mix-up in Frequency Domaincited by 12
  12. 2025GRU: Mitigating the Trade-off between Unlearning and Retention for LLMscited by 11
  13. 2025MTL-UE: Learning to Learn Nothing for Multi-Task Learningcited by 11
  14. 2025Directional Gradient Projection for Robust Fine-Tuning of Foundation Modelscited by 10
  15. 2025Towards Model Resistant to Transferable Adversarial Examples via Trigger Activationcited by 10
  16. 2024Unlearnable Examples Detection via Iterative Filteringcited by 9
  17. 2025From Pretrain to Pain: Adversarial Vulnerability of Video Foundation Models Without Task Knowledgecited by 8
  18. 2024Towards Physical World Backdoor Attacks against Skeleton Action Recognitioncited by 7
  19. 2025Theoretical Insights in Model Inversion Robustness and Conditional Entropy Maximization for Collaborative Inference Systemscited by 6
  20. 2026Time Is All It Takes: Spike-Retiming Attacks on Event-Driven Spiking Neural Networkscited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026FeatDistill: A Feature Distillation Enhanced Multi-Expert Ensemble Framework for Robust AI-generated Image Detectioncited by 0
  2. 2026Towards Universal Physical Adversarial Attacks via a Joint Multi-Objective and Multi-Model Optimization Frameworkcited by 0
  3. 2026When Modalities Fail to Tango: Conformal Backdoor Detection in Multimodal Contrastive Learningcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Rethinking Robust Adversarial Concept Erasure in Diffusion Models
  2. 2025When Are Concepts Erased From Diffusion Models?
  3. 2025Rethinking the Vulnerability of Concept Erasure and a New Method
  4. 2024Pruning for Robust Concept Erasing in Diffusion Models
  5. 2025Robust Concept Erasure in Diffusion Models: A Theoretical Perspective on Security and Robustness
  6. 2026No Concept Escapes the Audit: Auditing-Aware Unlearning for Verifiable Concept Erasure in Diffusion Models
  7. 2026To Erase, or Not to Erase: Robust Training-Free Concept Erasure with Preservation aware Adaptive Ranked Subspace Expansion
  8. 2025A Comprehensive Survey on Concept Erasure in Text-to-Image Diffusion Models