Enregistré dans:
Détails bibliographiques
Auteurs principaux: Zhao, Yutian, Du, Chao, Zheng, Xiaosen, Pang, Tianyu, Lin, Min
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2510.14269
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866911214109458432
author Zhao, Yutian
Du, Chao
Zheng, Xiaosen
Pang, Tianyu
Lin, Min
author_facet Zhao, Yutian
Du, Chao
Zheng, Xiaosen
Pang, Tianyu
Lin, Min
contents Data attribution for generative models seeks to quantify the influence of individual training examples on model outputs. Existing methods for diffusion models typically require access to model gradients or retraining, limiting their applicability in proprietary or large-scale settings. We propose a nonparametric attribution method that operates entirely on data, measuring influence via patch-level similarity between generated and training images. Our approach is grounded in the analytical form of the optimal score function and naturally extends to multiscale representations, while remaining computationally efficient through convolution-based acceleration. In addition to producing spatially interpretable attributions, our framework uncovers patterns that reflect intrinsic relationships between training data and outputs, independent of any specific model. Experiments demonstrate that our method achieves strong attribution performance, closely matching gradient-based approaches and substantially outperforming existing nonparametric baselines. Code is available at https://github.com/sail-sg/NDA.
format Preprint
id arxiv_https___arxiv_org_abs_2510_14269
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Nonparametric Data Attribution for Diffusion Models
Zhao, Yutian
Du, Chao
Zheng, Xiaosen
Pang, Tianyu
Lin, Min
Machine Learning
Data attribution for generative models seeks to quantify the influence of individual training examples on model outputs. Existing methods for diffusion models typically require access to model gradients or retraining, limiting their applicability in proprietary or large-scale settings. We propose a nonparametric attribution method that operates entirely on data, measuring influence via patch-level similarity between generated and training images. Our approach is grounded in the analytical form of the optimal score function and naturally extends to multiscale representations, while remaining computationally efficient through convolution-based acceleration. In addition to producing spatially interpretable attributions, our framework uncovers patterns that reflect intrinsic relationships between training data and outputs, independent of any specific model. Experiments demonstrate that our method achieves strong attribution performance, closely matching gradient-based approaches and substantially outperforming existing nonparametric baselines. Code is available at https://github.com/sail-sg/NDA.
title Nonparametric Data Attribution for Diffusion Models
topic Machine Learning
url https://arxiv.org/abs/2510.14269