AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

IdealCache: Rethinking Cache Scheduling in Diffusion Transformers via Ideal Trajectories

shang xue, Li Zhu, Ping Chen

NeurIPS 2026 · Image generation › Fast diffusion sampling · approach: Training-free diffusion acceleration · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes IdealCache, a feature-caching scheme for Diffusion Transformers that targets an ideal trajectory rather than the standard reference denoising trajectory to schedule cache reuse.

Problem

Feature caching speeds up Diffusion Transformers by reusing computations across timesteps, but existing methods approximate the standard reference denoising trajectory, which caps output quality at that trajectory's quality.

Approach

Argues the standard reference trajectory is an arbitrary optimization target and instead schedules caching to approximate an "ideal trajectory," removing the implicit quality ceiling inherited from the reference.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use ideal-trajectory schedule (IdealCache B=17) instead of HyCa N=3 at similar speedQwen-Image caching · DrawBenchSSIM: 0.763 -> 0.830Improved
use IdealCache B=17 instead of HyCa N=3Qwen-Image caching · DrawBenchImageReward: 1.1914 -> 1.1767Hurt
use IdealCache B=7 instead of HyCa N=8 at high compressionQwen-Image caching · DrawBenchImageReward: 1.0419 -> 1.1068Improved
use IdealCache B=9 instead of HyCa N=6 on image editingQwen-Image-Edit caching · GEdit-EN FullOverall Score: 7.45 -> 7.62Improved
IdealCache B=9 versus unaccelerated 50-step pipelineQwen-Image-Edit · GEdit-EN FullOverall Score: 7.54 -> 7.62Improved
set source image as ideal target for editing schedule (B=25)IdealCache on Qwen-Image-Edit · FlowEditBackground Preservation (SSIM): 0.665 -> 0.687Improved
IdealCache B=10 versus vanilla 50-step on FlowEditQwen-Image-Edit · FlowEditImageReward: 1.054 -> 0.991Hurt
IdealCache B=7 versus vanilla 50-step on text-to-imageQwen-Image · DrawBenchImageReward: 1.2092 -> 1.1068Hurt
build schedule graph from one prompt (R=1) instead of R=1000 ensemble (ablation)IdealCache · DrawBench ImageReward (B=9)ImageReward: 1.181 -> 1.168No clear effect
build schedule graph from one prompt (R=1) on GEdit-EN (ablation)IdealCache · GEdit-EN (B=9)Overall Score: 7.62 -> 7.58No clear effect
densify candidate grid from M=50 to M=100 (ablation)IdealCache · DrawBench 1664x928 (B=10)SSIM: 0.75 -> 0.78Improved
densify candidate grid from M=50 to M=100 (ablation)IdealCache · DrawBench 1664x928 (B=10)LPIPS: 0.33 -> 0.27Improved

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