Comprehensive Evaluation and Analysis for NSFW Concept Erasure in Text-to-Image Diffusion Models
On the map: Diffusion models → Robustness, privacy and adversarial aspects
Abstract (arXiv)
Text-to-image diffusion models have gained widespread application across various domains, demonstrating remarkable creative potential. However, the strong generalization capabilities of diffusion models can inadvertently lead to the generation of not-safe-for-work (NSFW) content, posing significant risks to their safe deployment. While several concept erasure methods have been proposed to mitigate the issue associated with NSFW content, a comprehensive evaluation of their effectiveness across various scenarios remains absent. To bridge this gap, we introduce a full-pipeline toolkit specifically designed for concept erasure and conduct the first systematic study of NSFW concept erasure methods. By examining the interplay between the underlying mechanisms and empirical observations, we provide in-depth insights and practical guidance for the effective application of concept erasure methods in various real-world scenarios, with the aim of advancing the understanding of content safety in diffusion models and establishing a solid foundation for future research and development in this critical area.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Robustness, privacy and adversarial aspects, one of 14 directions in Diffusion models. Builds on 39 of its 48 references that are in the map. First posted on arXiv 21 May 2025.
Builds on
39 of 48 references.
Closest ancestors · 21
Same direction, newest first
- 2025ACE: Anti-Editing Concept Erasure in Text-to-Image Models3 Jan
- 2024Meta-Unlearning on Diffusion Models: Preventing Relearning Unlearned Concepts16 Oct
- 2024Holistic Unlearning Benchmark: A Multi-Faceted Evaluation for Text-to-Image Diffusion Model Unlearning8 Oct
- 2024Growth Inhibitors for Suppressing Inappropriate Image Concepts in Diffusion Models2 Aug
- 2024Defensive Unlearning with Adversarial Training for Robust Concept Erasure in Diffusion Models24 May
- 2024MACE: Mass Concept Erasure in Diffusion Models10 Mar
- 2024UnlearnCanvas: Stylized Image Dataset for Enhanced Machine Unlearning Evaluation in Diffusion Models19 Feb
- 2024Erasing Undesirable Influence in Diffusion Models11 Jan
- 2023One-Dimensional Adapter to Rule Them All: Concepts, Diffusion Models and Erasing Applications26 Dec
- 2023SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation19 Oct
- 2023To Generate or Not? Safety-Driven Unlearned Diffusion Models Are Still Easy To Generate Unsafe Images ... For Now18 Oct
- 2023Ring-A-Bell! How Reliable are Concept Removal Methods for Diffusion Models?16 Oct
9 more
- 2023Prompting4Debugging: Red-Teaming Text-to-Image Diffusion Models by Finding Problematic Prompts12 Sep
- 2023Unified Concept Editing in Diffusion Models25 Aug
- 2023On the Trustworthiness Landscape of State-of-the-art Generative Models: A Survey and Outlook31 Jul
- 2023Unsafe Diffusion: On the Generation of Unsafe Images and Hateful Memes From Text-To-Image Models23 May
- 2023Forget-Me-Not: Learning to Forget in Text-to-Image Diffusion Models30 Mar
- 2023Ablating Concepts in Text-to-Image Diffusion Models23 Mar
- 2023Erasing Concepts from Diffusion Models13 Mar
- 2022Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Models9 Nov
- 2022Red-Teaming the Stable Diffusion Safety Filter3 Oct
Other references · 18
- 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
- 2020Denoising Diffusion Probabilistic Modelscited by 21,076
- 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
- 2014Microsoft COCO: Common Objects in Contextcited by 17,608
- 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,380
- 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,177
- 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,058
- 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 6,355
- 2022Classifier-Free Diffusion Guidancecited by 5,847
- 2022LAION-5B: An open large-scale dataset for training next generation image-text modelscited by 4,152
- 2021GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Modelscited by 3,614
- 2018Datasheets for Datasetscited by 1,386
6 more
- 2023ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generationcited by 1,045
- 2024Evaluating Text-to-Visual Generation with Image-to-Text Generationcited by 474
- 2018DeepFakes: a New Threat to Face Recognition? Assessment and Detectioncited by 209
- 2020Improving Image Captioning with Better Use of Captionscited by 84
- 2020Raiders of the Lost Kek: 3.5 Years of Augmented 4chan Posts from the Politically Incorrect Boardcited by 24
- 2024TextToon: Real-Time Text Toonify Head Avatar from Single Videocited by 21
Built on by
Citing papers per year: 2026: 1.
1 paper.
Similar, not cited
Close in wording, no citation either way
- 2025Comprehensive Assessment and Analysis for NSFW Content Erasure in Text-to-Image Diffusion Models
- 2025CE-SDWV: Effective and Efficient Concept Erasure for Text-to-Image Diffusion Models via a Semantic-Driven Word Vocabulary
- 2024Six-CD: Benchmarking Concept Removals for Benign Text-to-image Diffusion Models
- 2025A Comprehensive Survey on Concept Erasure in Text-to-Image Diffusion Models
- 2024Reliable and Efficient Concept Erasure of Text-to-Image Diffusion Models
- 2025Beautiful Images, Toxic Words: Understanding and Addressing Offensive Text in Generated Images
- 2025M-ErasureBench: A Comprehensive Multimodal Evaluation Benchmark for Concept Erasure in Diffusion Models
- 2025SuMa: A Subspace Mapping Approach for Robust and Effective Concept Erasure in Text-to-Image Diffusion Models