Metraipolis › archive › dc_4a46c435fa9b
RASteer: Retain-Aware Activation Steering for Concept Erasure in Diffusion Models — arXiv:2610.01969
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.01969
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-06T19:15:10.473Z
- SHA-256 of the archived text
58b0015ce718e85129fae9434bca75ed3b3dc71a62e3f14a987c2828b8e8116f- Stable URL of this page
https://metraipolis.com/documento/dc_4a46c435fa9b
How to cite
RASteer: Retain-Aware Activation Steering for Concept Erasure in Diffusion Models — arXiv:2610.01969 — arXiv (en), https://arxiv.org/abs/2610.01969 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4a46c435fa9b (Metraipolis archive dc_4a46c435fa9b, sha256 58b0015ce718e85129fae9434bca75ed3b3dc71a62e3f14a987c2828b8e8116f)
BibTeX
@misc{dc4a46c435fa9b,
title = {RASteer: Retain-Aware Activation Steering for Concept Erasure in Diffusion Models — arXiv:2610.01969},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.01969},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4a46c435fa9b; sha256 58b0015ce718e85129fae9434bca75ed3b3dc71a62e3f14a987c2828b8e8116f},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.01969. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Concept erasure aims to remove a target concept, such as a copyrighted style, a recognizable character, or unsafe content, from a pretrained text-to-image diffusion model while preserving its ability to generate other content. Existing activation steering methods build an erasure direction mainly from the target concept and adjust model activations along it at inference time. However, target and retained concepts often overlap in the model's representation space, so this direction also contains shared components that retained concepts rely on. Steering directly along this direction can therefore suppress retained concepts and harm the generation of non-target content. To address this issue, we propose Retain-aware Activation Steering (RASteer), a training-free method. RASteer first builds a retain subspace from the concepts to preserve. Retain-Orthogonal Steering (ROS) then removes components aligned with this subspace from the erasure direction, making steering more specific to the target. Since fully removing the shared components can weaken erasure, we further introduce Overlap-Adaptive Calibration (OAC). At each layer and denoising step, OAC uses the overlap between the erasure direction and the retain subspace to control how much of each shared component is removed, balancing target erasure and concept preservation. Experiments on unsafe-content, instance, and artistic-style erasure across multiple backbones and benchmarks show that RASteer matches or outperforms the activation steering and weight editing baselines we evaluate, achieving a better balance between erasure and preservation. Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-10-02 in cs.CV). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2610.01969 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-06T17:37:36.867Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-10-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).