AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

UniGP: Taming Diffusion Transformer for Prior-Preserved Unified Generation and Perception

Qin Guo, Hao Luo, Dongxu Yue, Weixuan Jin, Xiao Fu, Fan Wang, Dan Xu

ECCV 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 perception training to controllable generation (ablation)UniGP · Depth-to-imageRMSE to condition: 10.73 -> 6.89Improved
add generation training to perception (ablation)UniGP · NYUv2 depthAbsRel: 5.5 -> 5.2Improved
add generation training to perception (ablation)UniGP · NYUv2 normalmean angular error: 17.1 -> 16.4Improved
use binary-weighted unified training strategy (ablation)UniGP · COCO-5K text-to-imageFID: 25.59 -> 21.05Improved
use full data instead of 50% generation data (ablation)UniGP · Depth-to-imageRMSE to condition: 8.40 -> 6.89Improved
use full perception data instead of 50% (ablation)UniGP · NYUv2 depthAbsRel: 5.9 -> 5.2Improved
copy only the image branch (DUGP) instead of duplicating backbone (ablation)UniGP · Depth-to-imageRMSE to condition: 8.99 -> 6.89Improved
freeze backbone and copy branch instead of fine-tuning backbone (Marigold-style) (ablation)UniGP · COCO-5K text-to-imageFID: 25.71 -> 21.05Improved
balance control and perception layers at (12,12) instead of (6,18) (ablation)UniGP · Depth-to-imageRMSE to condition: 9.21 -> 6.89Improved
balance layers at (12,12) instead of (18,6) (ablation)UniGP · NYUv2 depthAbsRel: 5.9 -> 5.2Improved
duplicate image branchUniGP · Inference cost20-step inference time: 5.07s -> 9.54sHurt
use UniGP versus DepthAnything on NYUv2 depthUniGP · NYUv2 depthAbsRel: 4.3 -> 5.2Hurt

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