Metraipolis › archive › dc_cdda0292adeb

Enabling Preference-driven Unlearning in Few-step Distilled Text-to-Image Diffusion Models — arXiv:2610.10859

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.10859
License (as archived)
CC0-1.0
Archived
2026-10-09T06:23:54.368Z
SHA-256 of the archived text
4dd1ff81895f041cd2067c483556a2212e015b93efae8902354e4702f1d78bb1
Stable URL of this page
https://metraipolis.com/documento/dc_cdda0292adeb

How to cite

Enabling Preference-driven Unlearning in Few-step Distilled Text-to-Image Diffusion Models — arXiv:2610.10859 — arXiv (en), https://arxiv.org/abs/2610.10859 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_cdda0292adeb (Metraipolis archive dc_cdda0292adeb, sha256 4dd1ff81895f041cd2067c483556a2212e015b93efae8902354e4702f1d78bb1)
BibTeX
@misc{dccdda0292adeb,
  title = {Enabling Preference-driven Unlearning in Few-step Distilled Text-to-Image Diffusion Models — arXiv:2610.10859},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.10859},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_cdda0292adeb; sha256 4dd1ff81895f041cd2067c483556a2212e015b93efae8902354e4702f1d78bb1},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.10859. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Text-to-image diffusion models are increasingly distilled into few-step variants and being deployed to enable fast inference. However, their ability to generate harmful or undesired content poses significant safety risks. Data-driven unlearning methods suppress targeted generations by fine-tuning model weights using specialized unlearning objectives. Crucially, these objectives implicitly rely on multi-step denoising dynamics, an assumption that breaks down for few-step distilled (FSD) models, resulting in ineffective forgetting. Furthermore, performing unlearning on the non-distilled base model and subsequently re-distilling it to obtain an unlearned FSD model incurs substantial computational and time overhead, making it impractical in many settings. Hence, we address this limitation with a preference-driven unlearning framework that revisits Direct Preference Optimization (DPO) for diffusion models. We show that standard DPO and its unlearning derivatives, formulated around noise-prediction error, transfer poorly to FSD models due to their altered generation dynamics. To overcome this, we introduce a modified preference optimization formulation explicitly aligned with the few-step generation properties, enabling direct concept removal in FSD models while preserving few-step efficiency and maintaining strong retention of desirable (non-targeted) capabilities. We evaluate our framework primarily on identity and NSFW (nudity) removal tasks and also extend our method to object-level unlearning. Extensive experiments demonstrate consistent and effective forgetting, and strong retention performance, establishing our method as a practical and principled solution for unlearning in FSD models.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.10859 — 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 2026-10-09T05:50:31.900Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).