CE-SDWV: Effective and Efficient Concept Erasure for Text-to-Image Diffusion Models via a Semantic-Driven Word Vocabulary
International Journal of Computer Vision · 2026-07 · 134 (7)
DOI: 10.1007/s11263-026-02897-x ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Computer Vision, online erschienen am 08.07.2026. Verfasst von Jiahang Tu und 6 weiteren Autor:innen. Die Arbeit nennt 89 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Natural Science Foundation of Zhejiang Province.
Abstract (Originalsprache)
Large-scale text-to-image (T2I) diffusion models have achieved remarkable generative performance about various concepts. With the limitation of privacy and safety in practice, the generative capability concerning not-safe-for-work (NSFW) concepts is undesirable, e.g., producing sexually explicit photos, and licensed images. The concept erasure task for T2I diffusion models has attracted considerable attention and requires an effective and efficient method. To achieve this goal, we propose a CE-SDWV framework, which removes the target concepts (e.g., NSFW concepts) of T2I diffusion models in the text semantic space by only adjusting the text tokens and does not need to re-train the original model’s weights. Specifically, our framework first builds a target concept-related word vocabulary to enhance the representation of the target concepts within the text semantic space. We then utilize an adaptive semantic component suppression strategy to ablate the target concept-related semantic information in the text tokens. To further adapt the above text tokens to the original image semantic space, we propose an end-to-end gradient-orthogonal token optimization strategy. Extensive experiments on I2P and UnlearnCanvas benchmarks demonstrate the effectiveness and efficiency of our method. Code is available at https://github.com/TtuHamg/CE-SDWV.
Förderung
- National Natural Science Foundation of China
- Natural Science Foundation of Zhejiang Province
Zitieren
APA
Tu, J., Feng, Q., Dong, J., Zhao, H., Zhang, C., Sebe, N., & Qian, H. (2026). CE-SDWV: Effective and Efficient Concept Erasure for Text-to-Image Diffusion Models via a Semantic-Driven Word Vocabulary. International Journal of Computer Vision, 134(7). https://doi.org/10.1007/s11263-026-02897-x
Vancouver
Tu J, Feng Q, Dong J, Zhao H, Zhang C, Sebe N, et al. CE-SDWV: Effective and Efficient Concept Erasure for Text-to-Image Diffusion Models via a Semantic-Driven Word Vocabulary. International Journal of Computer Vision. 2026;134(7). doi:10.1007/s11263-026-02897-x
BibTeX
@article{Tu2026,
author = {Tu, Jiahang and Feng, Qian and Dong, Jiahua and Zhao, Hanbin and Zhang, Chao and Sebe, Nicu and Qian, Hui},
title = {CE-SDWV: Effective and Efficient Concept Erasure for Text-to-Image Diffusion Models via a Semantic-Driven Word Vocabulary},
journal = {International Journal of Computer Vision},
year = {2026},
volume = {134},
number = {7},
doi = {10.1007/s11263-026-02897-x}
}
EUVIMED