AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

Head-wise Adaptive Rotary Positional Encoding for Fine-Grained Image Generation

Jiaye Li, Baoyou Chen, Hui Li, Zilong Dong, Jingdong Wang, Siyu Zhu

CVPR 2026 · Image generation › Text-to-image diffusion · approach: Layout-to-image control

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
replace RoPE with HARoPEDiT-B/2 · ImageNet 256FID-50K: 9.49 -> 8.90Improved
make the adaptation matrix head-specific (ablation)RoPE + SVD (DiT-B/2) · ImageNet 256FID-50K: 8.93 -> 8.90Improved
make normal matrix head-specific (ablation)RoPE + normal-matrix · ImageNet 256FID-50K: 9.28 -> 9.03Improved
use SVD parameterization instead of normal or orthogonal matrix (ablation)HARoPE · ImageNet 256FID-50K: 9.28 / 9.31 -> 8.93Improved
add head-wise SVD rather than parameter-matched shared RoPE+SVD (ablation)FLUX.1-dev · GenEval Overall 2048GenEval overall: 0.731 -> 0.756Improved
add shared RoPE+SVD (no head-wise) (ablation)FLUX.1-dev · GenEval Overall 2048GenEval overall: 0.735 -> 0.731No clear effect
tune RoPE theta_base per frequency instead of HARoPE (ablation)DiT-B/2 100k steps · ImageNet 256FID: 20.77 (10,000) vs 20.45 (5,000) vs 19.01 HARoPENo clear effect
replace RoPE with HARoPE at 512x512FLUX.1-dev · GenEval Overall 512GenEval overall: 0.765 -> 0.758Hurt
replace RoPE with HARoPE at 1024x1024FLUX.1-dev · GenEval Overall 1024GenEval overall: 0.757 -> 0.771Improved
replace RoPE with HARoPE at 2048x2048FLUX.1-dev · GenEval PositionPosition score: 0.370 -> 0.455Improved
replace RoPE with HARoPEFLUX.1-dev · T2I-CompBench 2D-Spatial 1024score: 0.307 -> 0.337Improved
replace RoPE with HARoPESD3-medium · MS-COCO 256FID: 5.35 -> 5.22Improved

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