Model Reveals What to Cache: Profiling-Based Feature Reuse for Video Diffusion Models

Xuran MA, Yexin LIU, Yaofu LIU, Xianfeng WU, Mingzhe ZHENG, Zihao WANG, Ser-Nam LIM^*, Chao YANG^*

^*Corresponding author for this work

Research output: Contribution to conference › Conference Paper › peer-review

Original language	English
Publication status	Published - Oct 2025
Event	International Conference on Computer Vision (ICCV 2025) - Duration: 1 Oct 2025 → 1 Oct 2025

Conference	International Conference on Computer Vision (ICCV 2025)
Period	1/10/25 → 1/10/25

Cite this

@conference{7cc8aa7b6266454ab443c02cd9ed081b,

title = "Model Reveals What to Cache: Profiling-Based Feature Reuse for Video Diffusion Models",

author = "Xuran MA and Yexin LIU and Yaofu LIU and Xianfeng WU and Mingzhe ZHENG and Zihao WANG and Ser-Nam LIM and Chao YANG",

year = "2025",

month = oct,

language = "English",

note = "International Conference on Computer Vision (ICCV 2025) ; Conference date: 01-10-2025 Through 01-10-2025",

}

TY - CONF

T1 - Model Reveals What to Cache: Profiling-Based Feature Reuse for Video Diffusion Models

AU - MA, Xuran

AU - LIU, Yexin

AU - LIU, Yaofu

AU - WU, Xianfeng

AU - ZHENG, Mingzhe

AU - WANG, Zihao

AU - LIM, Ser-Nam

AU - YANG, Chao

PY - 2025/10

Y1 - 2025/10

M3 - Conference Paper

T2 - International Conference on Computer Vision (ICCV 2025)

Y2 - 1 October 2025 through 1 October 2025

ER -