HiCache: A Plug-in Scaled-Hermite Upgrade for Taylor-Style Cache-then-Forecast Diffusion Acceleration

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Feng, Liang, Zheng, Shikang, Liu, Jiacheng, Lin, Yuqi, Zhou, Qinming, Cai, Peiliang, Wang, Xinyu, Chen, Junjie, Zou, Chang, Ma, Yue, Zhang, Linfeng
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866908789012168704
author Feng, Liang
Zheng, Shikang
Liu, Jiacheng
Lin, Yuqi
Zhou, Qinming
Cai, Peiliang
Wang, Xinyu
Chen, Junjie
Zou, Chang
Ma, Yue
Zhang, Linfeng
author_facet Feng, Liang
Zheng, Shikang
Liu, Jiacheng
Lin, Yuqi
Zhou, Qinming
Cai, Peiliang
Wang, Xinyu
Chen, Junjie
Zou, Chang
Ma, Yue
Zhang, Linfeng
contents Diffusion models have achieved remarkable success in content generation but often incur prohibitive computational costs due to iterative sampling. Recent feature caching methods accelerate inference via temporal extrapolation, yet can suffer quality degradation from inaccurate modeling of the complex dynamics of feature evolution. We propose HiCache (Hermite Polynomial-based Feature Cache), a training-free acceleration framework that improves feature prediction by aligning mathematical tools with empirical properties. Our key insight is that feature-derivative approximations in diffusion Transformers exhibit multivariate Gaussian characteristics, motivating the use of Hermite polynomials as a potentially optimal basis for Gaussian-correlated processes. We further introduce a dual-scaling mechanism that ensures numerical stability while preserving predictive accuracy, and is also effective when applied standalone or integrated with TaylorSeer. Extensive experiments demonstrate HiCache's superiority, achieving 5.55x speedup on FLUX.1-dev while matching or exceeding baseline quality, and maintaining strong performance across text-to-image, video generation, and super-resolution tasks. Moreover, HiCache can be naturally added to previous caching methods to enhance their performance, e.g., improving ClusCa from 0.9480 to 0.9840 in terms of image rewards. Code: https://github.com/fenglang918/HiCache
format Preprint
id arxiv_https___arxiv_org_abs_2508_16984
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle HiCache: A Plug-in Scaled-Hermite Upgrade for Taylor-Style Cache-then-Forecast Diffusion Acceleration
Feng, Liang
Zheng, Shikang
Liu, Jiacheng
Lin, Yuqi
Zhou, Qinming
Cai, Peiliang
Wang, Xinyu
Chen, Junjie
Zou, Chang
Ma, Yue
Zhang, Linfeng
Computer Vision and Pattern Recognition
Diffusion models have achieved remarkable success in content generation but often incur prohibitive computational costs due to iterative sampling. Recent feature caching methods accelerate inference via temporal extrapolation, yet can suffer quality degradation from inaccurate modeling of the complex dynamics of feature evolution. We propose HiCache (Hermite Polynomial-based Feature Cache), a training-free acceleration framework that improves feature prediction by aligning mathematical tools with empirical properties. Our key insight is that feature-derivative approximations in diffusion Transformers exhibit multivariate Gaussian characteristics, motivating the use of Hermite polynomials as a potentially optimal basis for Gaussian-correlated processes. We further introduce a dual-scaling mechanism that ensures numerical stability while preserving predictive accuracy, and is also effective when applied standalone or integrated with TaylorSeer. Extensive experiments demonstrate HiCache's superiority, achieving 5.55x speedup on FLUX.1-dev while matching or exceeding baseline quality, and maintaining strong performance across text-to-image, video generation, and super-resolution tasks. Moreover, HiCache can be naturally added to previous caching methods to enhance their performance, e.g., improving ClusCa from 0.9480 to 0.9840 in terms of image rewards. Code: https://github.com/fenglang918/HiCache
title HiCache: A Plug-in Scaled-Hermite Upgrade for Taylor-Style Cache-then-Forecast Diffusion Acceleration
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2508.16984