Prototype-Guided Concept Erasure in Diffusion Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914379893571584 |
|---|---|
| author | Cai, Yuze Lu, Jiahao Shi, Hongxiang Zhou, Yichao Lu, Hong |
| author_facet | Cai, Yuze Lu, Jiahao Shi, Hongxiang Zhou, Yichao Lu, Hong |
| contents | Concept erasure is extensively utilized in image generation to prevent text-to-image models from generating undesired content. Existing methods can effectively erase narrow concepts that are specific and concrete, such as distinct intellectual properties (e.g. Pikachu) or recognizable characters (e.g. Elon Musk). However, their performance degrades on broad concepts such as ``sexual'' or ``violent'', whose wide scope and multi-faceted nature make them difficult to erase reliably. To overcome this limitation, we exploit the model's intrinsic embedding geometry to identify latent embeddings that encode a given concept. By clustering these embeddings, we derive a set of concept prototypes that summarize the model's internal representations of the concept, and employ them as negative conditioning signals during inference to achieve precise and reliable erasure. Extensive experiments across multiple benchmarks show that our approach achieves substantially more reliable removal of broad concepts while preserving overall image quality, marking a step towards safer and more controllable image generation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_08271 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Prototype-Guided Concept Erasure in Diffusion Models Cai, Yuze Lu, Jiahao Shi, Hongxiang Zhou, Yichao Lu, Hong Computer Vision and Pattern Recognition I.4.10; K.4.2 Concept erasure is extensively utilized in image generation to prevent text-to-image models from generating undesired content. Existing methods can effectively erase narrow concepts that are specific and concrete, such as distinct intellectual properties (e.g. Pikachu) or recognizable characters (e.g. Elon Musk). However, their performance degrades on broad concepts such as ``sexual'' or ``violent'', whose wide scope and multi-faceted nature make them difficult to erase reliably. To overcome this limitation, we exploit the model's intrinsic embedding geometry to identify latent embeddings that encode a given concept. By clustering these embeddings, we derive a set of concept prototypes that summarize the model's internal representations of the concept, and employ them as negative conditioning signals during inference to achieve precise and reliable erasure. Extensive experiments across multiple benchmarks show that our approach achieves substantially more reliable removal of broad concepts while preserving overall image quality, marking a step towards safer and more controllable image generation. |
| title | Prototype-Guided Concept Erasure in Diffusion Models |
| topic | Computer Vision and Pattern Recognition I.4.10; K.4.2 |
| url | https://arxiv.org/abs/2603.08271 |