AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

FADE: Fractional Anomalous Dynamics Extrapolation for Training-Free Diffusion Transformers Acceleration

Jinlong Yang, Jinke Wu, Lizilin, Yao Zhou

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 FADE, a training-free feature-caching method for accelerating Diffusion Transformer sampling by forecasting features with fractional-order (anomalous dynamics) extrapolation instead of integer-order operators.

Problem

Iterative sampling in Diffusion Transformers is computationally expensive; existing feature-caching methods that reuse or forecast features with integer-order operators adapt poorly to the feature dynamics across denoising steps.

Approach

Replaces integer-order difference/extrapolation operators used in feature-caching forecasters with fractional-order extrapolation modeling anomalous (non-integer-order) feature dynamics, requiring no retraining.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use stage-aware LUT-calibrated fractional order instead of integer order gamma=1 (ablation)FADE · DiT-XL/2 FID, N=6FID: 3.17 -> 2.87Improved
use fixed global fractional order 0.85 instead of gamma=1 (ablation)FADE · DiT-XL/2 FID, N=6FID: 3.17 -> 3.06Improved
use stage-aware LUT instead of polynomial fit of gamma(t) (ablation)FADE · DiT-XL/2 FID, N=6FID: 3.00 -> 2.87Improved
calibrate with directional cosine loss instead of MSE (ablation)FADE · DiT-XL/2 FID, N=6FID: 2.97 -> 2.87Improved
use FADE instead of HiCache at same costFADE · DiT-XL/2 FID, N=4FID: 2.51 -> 2.41Improved
use FADE instead of FORA at N=6FADE · DiT-XL/2 FID, N=6FID: 9.22 -> 2.87Improved
increase calibration set size from 1 to 10 samples (ablation)FADE · FLUX.1 ImageReward, N=6ImageReward: 0.9963 -> 1.0019Improved
increase calibration set size from 10 to 20 (ablation)FADE · FLUX.1 ImageReward, N=6ImageReward: 1.0019 -> 1.0008No clear effect
add global refinement factor lambda=1.3 (ablation)FADE · FLUX.1 ImageReward, N=5ImageReward: 0.9974 -> 1.0071Improved
increase lambda from 1.3 to 1.4 (ablation)FADE · FLUX.1 ImageReward, N=5ImageReward: 1.0071 -> 0.9858Hurt
use FADE instead of HiCache at N=7 m=1FADE · FLUX.1 PSNR, N=7PSNR: 18.2778 (HiCache) vs 18.1925 (FADE)Mixed
use FADE instead of TaylorSeer on HunyuanVideoFADE · HunyuanVideo VBench, N=6VBench: 79.87 -> 80.12Improved

Experimental setup

Builds on: feature caching for diffusion transformers

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