RaPD: Resolution-Agnostic Pixel Diffusion via Semantics-Enriched Implicit Representations

Yanhao Ge, Shanyan Guan, Weihao Wang, Ying Tai, Mingyu You · 2026

Read RaPD: Resolution-Agnostic Pixel Diffusion via Semantics-Enriched Implicit Representations by Yanhao Ge, Shanyan Guan, Weihao Wang, Ying Tai, Mingyu You (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-15908.txt.

Description

Yanhao Ge and four colleagues propose RaPD, size-free pixel diffusion in a Neural Image Field latent space. Natural images are smooth, yet most models draw them on grids, blocking size-free output. Prior neural fields add smoothness only at decode time via interpolation, leaving the latent space gridded. RaPD bridges this gap with Semantic Representation Guidance for learning generation-ready latents and a Coordinate-Queried Attention Renderer for scale-aware output at chosen coordinates. One denoised latent renders at any size by changing query points alone, keeping cost fixed.

Catalog record

CREATORYanhao Ge, Shanyan Guan, Weihao Wang, Ying Tai, Mingyu You
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE48719
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.15908

Browse the full library · About the Tower of Babel