Computer-Aided Design Generation by Cascaded Discrete Diffusion Model
Honghu Pan, Xiaoling Luo, Yongyong Chen, Member, IEEE, Zhenyu He, Senior member, IEEE, Pengyang Wang · 2026
Read Computer-Aided Design Generation by Cascaded Discrete Diffusion Model by Honghu Pan, Xiaoling Luo, Yongyong Chen, Member, IEEE, Zhenyu He, Senior member, IEEE, Pengyang Wang (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-2605-05031.txt.
Description
Honghu Pan and four co-authors generate computer-aided design models with a diffusion process that respects the discrete nature of design commands. A CAD model is a sequence of commands and values, and smoothing those symbols in a smooth space often produces invalid ones. Their framework therefore diffuses directly over symbol classes instead.
One stage generates the commands, and a second generates the values tied to those commands. Special step rules handle where things sit, sizes, and yes-or-no traits in ways suited to each kind. The cascaded design aims to produce valid, editable CAD sequences more reliably than step-by-step or continuous diffusion approaches.
Catalog record
| CREATOR | Honghu Pan, Xiaoling Luo, Yongyong Chen, Member, IEEE, Zhenyu He, Senior member, IEEE, Pengyang Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.05031