arXiv：cs.LG（机器学习，全量分类）· Th\'eau d'Audiffret, Mariia Vladimirova, Jean-Yves Franceschi·· 14 小时前AI 评分43
无需监督的扩散模型公平性与多样性去偏方法
Debias Anything: Fairness with Diversity without Supervision in Diffusion Models
AI 导读
研究者提出一种可同时提升扩散模型公平性与多样性的方法，通过适配器将冻结的扩散模型连接到预训练视觉-语言嵌入空间，无需敏感属性标注即可实现引导。该方法用成对文本提示定义属性方向以控制批次构成比例，并引入语义估计分歧分数来促进多样性，适用于无条件与文本条件扩散模型。实验显示，该方法在公平性水平相当的情况下提升了生成质量与多样性分数。
正文
Abstract:Although diffusion models produce high-quality images, they also reproduce and amplify demographic imbalances in their training data. Debiasing their generation process post-training w.r.t. some sensitive attribute usually relies on classifier guidance or explicit text extra-conditioning, but this reduces methods' applicability and output diversity. Conversely, methods promoting diversity alone do not ensure fair attribute representation. In this paper, we propose a method tackling fairness and diversity jointly that is generally applicable to any diffusion model and any sensitive attribute. To this end, an adapter connects the frozen diffusion model to a pretrained vision-language embedding space, enabling fairness and diversity guidance without sensitive-attribute annotations. For fairness, pairs of text prompts define attribute directions which guide batch composition towards specific proportions. For diversity, we introduce a score measuring disagreement between the semantic estimates derived from this representation. The formulation supports unconditional and text-conditional diffusion models, while requiring no prior knowledge or data of sensitive attribute. Experiments confirm that our method improves quality and diversity scores at comparable fairness levels.
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.01815 [cs.LG] |
| (or arXiv:2610.01815v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.01815 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Mariia Vladimirova [view email]
[v1]
Thu, 1 Oct 2026 14:51:03 UTC (10,690 KB)
来源：arXiv：cs.LG（机器学习，全量分类） · arxiv.org