AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Phase-Aligned RoPE for Mixed-Resolution Diffusion Transformer

Haoyu Wu, Jingyi Xu, Qiaomu Miao, Dimitris Samaras, Hieu Le

ECCV 2026 · Image generation › Autoregressive and latent generation · approach: High-resolution diffusion synthesis · 2 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
use phase-aligned attention (PMA) instead of PI-LRRoPE position interpolation · Wan2.1-1.3B mixed-resolution RoPE interpolation comparisonVBench Total: 71.73 -> 77.04Improved
use PMA instead of PI-HRRoPE position interpolation · Wan2.1-1.3B mixed-resolution RoPE interpolation comparisonDOVER Overall: 35.04 -> 75.34Improved
use PMA instead of YARNRoPE interpolation · Wan2.1-1.3B mixed-resolution RoPE interpolation comparisonVBench Total: 72.46 -> 77.04Improved
use PMA instead of NTKRoPE interpolation · FLUX.1-dev mixed-resolution RoPE interpolation comparisonImageReward: 0.953 -> 0.978Improved
add Boundary Expand-and-Replace (n_pad 2/2) (ablation)PMA · Boundary band ablation (Wan2.1-1.3B)VBench Total: 74.74 -> 77.04Improved
add Boundary Expand-and-Replace (n_pad 2/2) (ablation)PMA · Boundary band ablation (Wan2.1-1.3B)DOVER Overall: 68.43 -> 75.34Improved
increase HR boundary padding from 2 to 4 (ablation)PMA · Boundary band ablation (Wan2.1-1.3B)VBench Total: 77.04 -> 76.90No clear effect
use Center Square instead of saliency model for HR regions (ablation)PMA · Wan2.1-1.3B mixed-resolution RoPE interpolation comparisonVBench Total: 77.04 -> 76.75No clear effect
use PMA instead of MagCache at 4xWan2.1-1.3B acceleration · Wan2.1-1.3B acceleration comparison (about 4x)VBench Total: 73.88 -> 77.04Improved
use PMA instead of ToCa at 6xFLUX.1-dev acceleration · FLUX.1-dev acceleration comparison (6x speedup)ImageReward: 0.345 -> 0.929Improved
combine PMA with MagCachePMA · Wan2.1-1.3B acceleration comparison (about 4x)Speedup: 4.0x -> 7.8xImproved
combine PMA with DMD 4-stepDMD (4-step) · Wan2.1-1.3B step distillationSpeedup: 15.8x -> 30.7xImproved

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