DreamPartGen : Semantically Grounded Part-Level 3D Generation via Collaborative Latent Denoising

Tianjiao Yu, Xinzhuo Li, Muntasir Wahed, Jerry Xiong, Yifan Shen, Ying Shen, Ismini Lourentzou {ty41,lourent2}@illinois.edu · 2026

Read DreamPartGen : Semantically Grounded Part-Level 3D Generation via Collaborative Latent Denoising by Tianjiao Yu, Xinzhuo Li, Muntasir Wahed, Jerry Xiong, Yifan Shen, Ying Shen, Ismini Lourentzou {ty41,lourent2}@illinois.edu (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-2603-19216.txt.

Description

Yu and co-authors build 3D objects part by part, with each part tied to words that name it. Most text-to-3D tools ignore what parts mean and how they relate. Their DreamPartGen uses paired latents for each part's shape and look. A second set of latents holds links between parts, drawn from language. A joint denoising step keeps shape and meaning in sync. On standard tests, the method leads in shape quality and text match.

Catalog record

CREATORTianjiao Yu, Xinzhuo Li, Muntasir Wahed, Jerry Xiong, Yifan Shen, Ying Shen, Ismini Lourentzou {ty41,lourent2}@illinois.edu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE61307
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.19216

Browse the full library · About the Tower of Babel