FllumaOne: A Code-Native Multimodal CAD Dataset with Executable Programs and Kernel-Validated Feature Histories

FllumaOne is a newly released multimodal CAD dataset designed for editable CAD research, consisting of 100,000 samples generated by executable Python programs within the Flluma CAD system. It includes a structured feature tree, STEP geometry, and various metadata formats, ensuring high fidelity with a baseline model, Qwen2.5-Coder-1.5B LoRA, achieving 99.98% Python syntax validity and 99.14% STEP-export validity on a test set. This dataset is significant for practitioners as it facilitates advanced tasks such as conditioned CAD reconstruction, executable program synthesis, and editable reverse engineering, enhancing the development of AI tools in CAD environments.