Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | https://arxiv.org/abs/2509.21008 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866912607715196928 |
|---|---|
| author | He, Qinqin Weng, Jiaqi Tao, Jialing Xue, Hui |
| author_facet | He, Qinqin Weng, Jiaqi Tao, Jialing Xue, Hui |
| contents | Text-to-image models exhibit remarkable capabilities in image generation. However, they also pose safety risks of generating harmful content. A key challenge of existing concept erasure methods is the precise removal of target concepts while minimizing degradation of image quality. In this paper, we propose Single Neuron-based Concept Erasure (SNCE), a novel approach that can precisely prevent harmful content generation by manipulating only a single neuron. Specifically, we train a Sparse Autoencoder (SAE) to map text embeddings into a sparse, disentangled latent space, where individual neurons align tightly with atomic semantic concepts. To accurately locate neurons responsible for harmful concepts, we design a novel neuron identification method based on the modulated frequency scoring of activation patterns. By suppressing activations of the harmful concept-specific neuron, SNCE achieves surgical precision in concept erasure with minimal disruption to image quality. Experiments on various benchmarks demonstrate that SNCE achieves state-of-the-art results in target concept erasure, while preserving the model's generation capabilities for non-target concepts. Additionally, our method exhibits strong robustness against adversarial attacks, significantly outperforming existing methods. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_21008 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | A Single Neuron Works: Precise Concept Erasure in Text-to-Image Diffusion Models He, Qinqin Weng, Jiaqi Tao, Jialing Xue, Hui Computer Vision and Pattern Recognition Text-to-image models exhibit remarkable capabilities in image generation. However, they also pose safety risks of generating harmful content. A key challenge of existing concept erasure methods is the precise removal of target concepts while minimizing degradation of image quality. In this paper, we propose Single Neuron-based Concept Erasure (SNCE), a novel approach that can precisely prevent harmful content generation by manipulating only a single neuron. Specifically, we train a Sparse Autoencoder (SAE) to map text embeddings into a sparse, disentangled latent space, where individual neurons align tightly with atomic semantic concepts. To accurately locate neurons responsible for harmful concepts, we design a novel neuron identification method based on the modulated frequency scoring of activation patterns. By suppressing activations of the harmful concept-specific neuron, SNCE achieves surgical precision in concept erasure with minimal disruption to image quality. Experiments on various benchmarks demonstrate that SNCE achieves state-of-the-art results in target concept erasure, while preserving the model's generation capabilities for non-target concepts. Additionally, our method exhibits strong robustness against adversarial attacks, significantly outperforming existing methods. |
| title | A Single Neuron Works: Precise Concept Erasure in Text-to-Image Diffusion Models |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2509.21008 |