Disentangled Sparse Representations for Concept-Separated Diffusion Unlearning
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_ | 1866916005017550848 |
|---|---|
| author | Kim, Hyeonjin Jung, Hangyeol Yun, Heechan Yun, Sungjun Han, Dong-Jun |
| author_facet | Kim, Hyeonjin Jung, Hangyeol Yun, Heechan Yun, Sungjun Han, Dong-Jun |
| contents | Unlearning specific concepts in text-to-image diffusion models has become increasingly important for preventing undesirable content generation. Among prior approaches, sparse autoencoder (SAE)-based methods have attracted attention due to their ability to suppress target concepts through lightweight manipulation of latent features, without modifying model parameters. However, SAEs trained with sparse reconstruction objectives do not explicitly enforce concept-wise separation, resulting in shared latent features across concepts. To address this, we propose SAEParate, which organizes latent representations into concept-specific clusters via a concept-aware contrastive objective, enabling more precise concept suppression while reducing unintended interference during unlearning. In addition, we enhance the encoder with a GeLU-based nonlinear transformation to increase its expressive capacity under this separation objective, enabling a more discriminative and disentangled latent space. Experiments on UnlearnCanvas demonstrate state-of-the-art performance, with particularly strong gains in joint style-object unlearning, a challenging setting where existing methods suffer from severe interference between target and non-target concepts. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_12122 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Disentangled Sparse Representations for Concept-Separated Diffusion Unlearning Kim, Hyeonjin Jung, Hangyeol Yun, Heechan Yun, Sungjun Han, Dong-Jun Machine Learning Artificial Intelligence Computer Vision and Pattern Recognition Unlearning specific concepts in text-to-image diffusion models has become increasingly important for preventing undesirable content generation. Among prior approaches, sparse autoencoder (SAE)-based methods have attracted attention due to their ability to suppress target concepts through lightweight manipulation of latent features, without modifying model parameters. However, SAEs trained with sparse reconstruction objectives do not explicitly enforce concept-wise separation, resulting in shared latent features across concepts. To address this, we propose SAEParate, which organizes latent representations into concept-specific clusters via a concept-aware contrastive objective, enabling more precise concept suppression while reducing unintended interference during unlearning. In addition, we enhance the encoder with a GeLU-based nonlinear transformation to increase its expressive capacity under this separation objective, enabling a more discriminative and disentangled latent space. Experiments on UnlearnCanvas demonstrate state-of-the-art performance, with particularly strong gains in joint style-object unlearning, a challenging setting where existing methods suffer from severe interference between target and non-target concepts. |
| title | Disentangled Sparse Representations for Concept-Separated Diffusion Unlearning |
| topic | Machine Learning Artificial Intelligence Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2605.12122 |