AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Group Diffusion: Enhancing Image Generation by Unlocking Cross-Sample Collaboration

Sicheng Mo, Thao Nguyen, Richard Zhang, Nick Kolkin, Siddharth Srinivasan Iyer, Eli Shechtman, Krishna Kumar Singh, Yong Jae Lee, Bolei Zhou, Yuheng Li

CVPR 2026 · Image generation › Autoregressive and latent generation · approach: High-resolution diffusion synthesis · 0 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 group attention on unconditional model only (C=1, UC=4, class query) (ablation)DiT-XL/2 · ImageNet 256x256FID (w/ CFG): 3.08 -> 2.81Improved
increase group size from 2 to 16 (CLIP-L) (ablation)GroupDiff DiT-XL/2 · ImageNet 256x256FID (w/ CFG): 2.92 -> 1.86Improved
replace random groups with CLIP-L similarity groups (ablation)GroupDiff DiT-XL/2 · ImageNet 256x256FID (w/ CFG): 3.57 -> 2.42Improved
use class-based instead of random groups (ablation)GroupDiff DiT-XL/2 · ImageNet 256x256FID (w/ CFG): 3.57 -> 2.81Improved
swap pretrained encoder for query (CLIP-L vs DINOv2-B) (ablation)GroupDiff DiT-XL/2 · ImageNet 256x256FID (w/ CFG): 2.42 vs 2.51No clear effect
add noise-level variation of 50 within group (ablation)GroupDiff DiT-XL/2 · ImageNet 256x256FID (w/ CFG): 2.42 -> 2.34Improved
apply group attention only in timesteps 0.0-0.4 (ablation)GroupDiff DiT-XL/2 · ImageNet 256x256FID-10K: 4.21 -> 3.92Improved
apply group attention up to timestep 0.6 (ablation)GroupDiff DiT-XL/2 · ImageNet 256x256FID-10K: 4.21 -> 4.63Hurt
keep group attention in layers 1-9 (ablation)GroupDiff DiT-XL/2 · ImageNet 256x256FID-10K: 294.38 -> 4.21Improved
keep group attention in layers 20-27 (ablation)GroupDiff DiT-XL/2 · ImageNet 256x256FID-10K: 4.49 -> 4.21No clear effect
use CLIP-L instead of class query when resuming SiT-XL/2 for 100 epochs (ablation)SiT-XL/2 GroupDiff-4* · ImageNet 256x256FID: 1.76 -> 1.40Improved
add GroupDiff-4 trained from scratch to DiT-XL/2DiT-XL/2 · ImageNet 256x256FID: 2.27 -> 1.66Improved

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