FllumaOne: A Code-Native Multimodal CAD Dataset with Executable Programs and Kernel-Validated Feature Histories
Researchers have introduced FllumaOne, a new dataset designed for editable CAD research. This dataset comprises 100,000 samples generated by executable Python programs within the Flluma CAD system. Each sample includes the program, a structured feature tree, STEP geometry, point cloud data, natural-language descriptions, and renderings. A baseline model trained on FllumaOne demonstrated high validity in syntax, build success, and export checks, indicating its potential for various CAD-related AI tasks. AI