AI Paper Atlas

Research map / Video / Video generation

Adaptive Caching for Faster Video Generation with Diffusion Transformers

Kumara Kahatapitiya, Haozhe Liu, Sen He, Ding Liu, Menglin Jia, Chenyang Zhang, Michael S. Ryoo, Tian Xie

ICCV 2025 · Video › Video generation · approach: Efficient video diffusion models · 73 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 AdaCache-fast (adaptive caching) to Open-SoraOpen-Sora 480p · Open-Sora 480p-2sspeedup: 1.00x -> 2.24xImproved
use AdaCache-fast instead of PAB-fastOpen-Sora-Plan · Open-Sora-Plan 512x512speedup: 1.45x -> 3.70xImproved
use AdaCache-fast instead of PAB-fastLatte · Latte 512x512speedup: 1.33x -> 2.74xImproved
add motion regularization to AdaCache-fast (ablation)AdaCache-fast · Open-Sora-Plan 512x512VBench (%): 75.83 -> 79.30Improved
add motion regularization to AdaCache-fast (ablation)AdaCache-fast · Open-Sora 480pVBench (%): 79.39 -> 79.48Improved
add motion regularization to AdaCache (ablation)AdaCache · Open-Sora 720p-2sVBench (%): 83.40 -> 83.50Improved
add motion-gradient as early motion predictor (ablation)AdaCache + MoReg · Open-Sora 720p-2sVBench (%): 83.36 -> 83.50Improved
estimate motion over multiple steps of frame differences (ablation)AdaCache + MoReg · Open-Sora 720p-2slatency (s): 93.50 -> 95.65Hurt
use cosine distance instead of L1 for the cache metric (ablation)AdaCache · Open-Sora 720p-2sVBench (%): 83.40 -> 83.19Hurt
compute cache metric at start instead of mid-DiT (ablation)AdaCache · Open-Sora 720p-2sVBench (%): 83.40 -> 83.30No clear effect
use MLP residual instead of temporal attention for cache metric (ablation)AdaCache · Open-Sora 720p-2slatency (s): 89.53 -> 99.72Hurt
use slow basis cache-rates instead of fast (ablation)AdaCache · Open-Sora 720p-2sVBench (%): 83.40 -> 84.12Improved

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