AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

ParaLin: Accelerating Parallel Diffusion Integrator via Intrinsic Partially Linear Structure

Jianrong Lu, Zhiyu Zhu, Hui LIU, Junhui Hou

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.

Accelerates diffusion sampling by reformulating sequential integral solvers as a parallel process exploiting an intrinsic partially linear structure (ParaLin).

Problem

Sequential inference of diffusion probabilistic models is slow because each denoising step depends on the previous one.

Approach

Views diffusion sampling as a dynamical system; shows the sequential integral solver can be approximated by a linear solver, enabling parallel sampling that exploits a partially linear structure.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use 16-step sequential UniPC/DPMSolver++ instead of 30-step EulerFLUX.1-dev · GenEvalGenEval: 0.66 -> 0.50-0.52Hurt
use 16-step sequential UniPC/DPMSolver++ instead of 30-step EulerFLUX.1-dev · DPG-BenchDPG-Bench: 83.97 -> 75.75-77.46Hurt
use 16-step sequential UniPC/DPMSolver++ instead of 30-step EulerFLUX.1-dev · LongTextBenchLongTextBench: 0.27 -> 0.08-0.09Hurt
keep 30 discrete steps and parallelize solver (ParaLin++)FLUX.1-dev · GenEval/DPG/LongTextBenchquality profile: 0.64 vs 0.66 GenEval, 84.24 vs 83.97 DPG, 0.26 vs 0.27 LTBNo clear effect
keep 30 discrete steps and parallelize solver (ParaLin++)Ming-Flash-Omni 2.0 · GenEvalGenEval: 0.88 -> 0.87No clear effect
increase number of sequential steps for few-step solvers (16 to 30) (ablation)UniPC on FLUX.1-dev · GenEvalGenEval: 0.52 -> 0.60Improved
vary CFG from 0 to 14 for 20-step few-step solvers (ablation)UniPC and DPMSolver++ on FLUX.1-dev · DPG-BenchDPG-Bench: 68.85-79.67 vs Euler 82.25-83.97Hurt
add parallel solver ParaLin++ to DDIMStable Diffusion-v2, 50 steps · wall-clock speedupspeedup: 1.0x -> 4.8xImproved
use full nonlinear parallel system instead of partial nonlinearity (ablation)ParaLin++ on 100-step DDIM · ablationNFE and speedup: partial nonlinearity removes 93.3% of NFEsHurt
add adaptive stopping criterion (ablation)ParaLin++ on 100-step DDIM · ablationiterations and NFEImproved
add successive over-relaxation (SOR) (ablation)ParaLin++ on 100-step DDIM · ablationspeedup: 15 iterations, 117 NFEs, 6.5x speedupImproved
run on a single GPUParaDiGMS vs ParaLin++ · Stable Diffusion-v2 DDIM-100speedup: ParaDiGMS 0.8x; ParaLin++ 4.2xMixed

Experimental setup

Builds on: DDIM Picard iteration ParaDiGMS

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