AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Pyramid Patchification Flow for Visual Generation

Hui Li, Baoyou Chen, Li jiaye, Jingdong Wang, Siyu Zhu

ICLR 2026 · Image generation › Autoregressive and latent generation · approach: High-resolution diffusion synthesis

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
use pyramidal patchification (larger patches at high noise), two stagesSiT-B/2 trained from scratch · ImageNet 256FID-50k: 4.46 -> 3.83 at 62.0 testing FLOPs (%)Improved
use three-stage pyramidal patchificationSiT-B/2 trained from scratch · ImageNet 256FID-50k: 4.46 -> 4.43No clear effect
use pyramid representations instead of full-resolution latents (ablation)PPFlow · ImageNet 256, 1M steps from scratchFID-50k: 15.68 -> 164.48Hurt
add renoising trick to pyramid representations (ablation)Pyramidal Flow variant · ImageNet 256, 1M stepsFID-50k: 164.48 -> 27.69Improved
train each patch size only on its own timestep range instead of all timesteps (Lumina-Video style) (ablation)PPFlow · ImageNet 256, 1M stepsFID-50k: 18.77 -> 15.68Improved
add learned patch-level (stage) embedding (ablation)PPF-B-2 · ImageNet 256, 1M steps from SiT-B/2FID-50k: 4.44 -> 4.30Improved
add stage-wise CFG schedule (ablation)PPF-B-2 · ImageNet 256, 1M steps from SiT-B/2FID-50k: 4.30 -> 4.22Improved
increase number of stages from 2 to 5 with equal 1M-step training (ablation)PPF-B · ImageNet 256, from SiT-B/2FID-50k: 4.22 -> 5.30Hurt
train five-stage PPFlow longer (ablation)PPF-B-5 · ImageNet 256, from SiT-B/2FID-50k: 5.30 -> 4.51Improved
move first stage boundary earlier (more full-resolution steps) (ablation)PPF-B-3 · ImageNet 256FID-50k / testing FLOPs: FID 4.57 -> 4.21; FLOPs 49.1 -> 68.3Mixed
apply PPFlow to FLUX.1-dev text-to-imageFLUX.1-ft · GenEval, DPG Bench, T2I-CompBench at 512 to 2048GenEval: 0.68 -> 0.68 at 62.2 FLOPs (%) (512)No clear effect
use PPFlow instead of FlexiDiT at about 63% FLOPsDiT-XL/2 · ImageNet 256FID-50k: 2.25 -> 2.15Improved

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