SK-Adapter: Skeleton-Based Structural Control for Native 3D Generation
Anbang Wang, Yuzhuo Ao, Shangzhe Wu, Chi-Keung Tang · 2026
Read SK-Adapter: Skeleton-Based Structural Control for Native 3D Generation by Anbang Wang, Yuzhuo Ao, Shangzhe Wu, Chi-Keung Tang (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-2603-14152.txt.
Description
Wang and co-authors add precise skeleton control to native 3D generation with SK-Adapter. Text and image prompts stay vague about exact pose and joint layout. Their light adapter turns joint positions and bone links into tokens that a frozen 3D backbone reads through cross-attention. The base model keeps its shape and texture quality while following the skeleton. A new dataset, Objaverse-TMS, gives 24,000 text, mesh, and skeleton triples for training. Tests beat prior baselines and extend to local 3D editing.
Catalog record
| CREATOR | Anbang Wang, Yuzhuo Ao, Shangzhe Wu, Chi-Keung Tang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 43957 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.14152