AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

OmniCache: A Trajectory-Oriented Global Perspective on Training-Free Cache Reuse for Diffusion Transformer Models

Huanpeng Chu, Wei Wu, Guanyu Feng, Yutao Zhang

ICCV 2025 · Image generation › Fast diffusion sampling · approach: Training-free diffusion acceleration · 12 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add noise correction module (ablation)OmniCache · CogVideoX-5b-I2V-distill I2VQ-Align: 0.778 -> 0.788Improved
add noise correction module (ablation)OmniCache · CogVideoX-5b-I2V-distill I2VAesthetic Quality: 0.58 -> 0.593Improved
add high-pass and low-pass filtering to noise estimate (ablation)OmniCache · CogVideoX-5b-I2V-distill I2VAesthetic Quality: 0.593 -> 0.621Improved
add high-pass and low-pass filtering to noise estimate (ablation)OmniCache · CogVideoX-5b-I2V-distill I2VQ-Align: 0.788 -> 0.792Improved
cache reuse alone on distilled model (ablation)OmniCache · CogVideoX-5b-I2V-distill I2VAesthetic Quality: 0.59 -> 0.58Hurt
use OmniCache instead of AdaCacheAdaCache · CogVideoX-5b-I2V-distill I2VQ-Align: 0.77 -> 0.792Improved
use OmniCache-slow instead of PAB-slowPAB-slow · Open-Sora VBenchVBench (%): 78.51 -> 78.83Improved
use OmniCache-fast instead of PAB-fastPAB-fast · Latte VBenchVBench (%): 73.13 -> 77.09Improved
use OmniCache-fast instead of TeaCache-fastTeaCache-fast · Latte VBenchVBench (%): 76.69 -> 77.09Mixed
add MoReg to AdaCache-fast (ablation)AdaCache-fast · Latte VBenchVBench (%): 76.26 -> 76.47Mixed
use OmniCache-slow on Open-SoraOpen-Sora · Open-Sora VBenchVBench (%): 79.22 -> 78.83No clear effect
apply OmniCache-fast to DiT-XL/2 class-conditionalDiT-XL/2 · ImageNet 256FID: 2.68 -> 2.69No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy