Skip to content

hamkerlab/Habibi2026_Tokenization_CADmodels

Repository files navigation

Habibi2026_Tokenization_CADmodels

Source code of the article Influence of Tokenization Strategies on the Prediction of CAD Model Descriptions by Sayeda Hadisa Habibi, Julia Bergelt, Michael Teichmann, Fred H. Hamker, accepted for the International Symposium on Hybrid Intelligence in Product and Production Engineering (2026).

Aim

We compare three different tokenization strategies:

on advanced text-based representations for CAD models.

We base our comparison of the different tokenization methods on the metrics token fertility and prediction performance on masked sequence parts, using a lightweight BERT-style transformer model and the DeepCAD dataset.

Getting Started

About

Source code of Habibi, S. H., Bergelt, J., Techmann, M. Hamker, F.H. (2026) Influence of Tokenization Strategies on the Prediction of CAD Model Descriptions. International Symposium on Hybrid Intelligence in Product and Production Engineering (accepted)

Resources

Stars

Watchers

Forks

Packages

 
 
 

Contributors

Languages