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

2026-06-16 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Data Science & Analytics · Depth: Expert, quick

Summary

FllumaOne is a new code-native multimodal CAD dataset designed for editable CAD research, featuring models generated by executable Python programs within the Flluma CAD system, which is based on Qt/C++ OpenCASCADE. Each dataset sample meticulously aligns its Python program with a structured feature tree, a training-oriented intermediate representation, STEP geometry, a surface point cloud, natural-language descriptions, metadata, and eight canonical visible-edge renderings. The primary release, FllumaOne-100K, comprises 100,000 validated samples across four template-level complexity regimes. Programs undergo rigorous kernel geometry, solid validity, and export checks before inclusion. A Qwen2.5-Coder-1.5B LoRA baseline, trained on 80,000 samples, demonstrated high performance on a 10,000-sample test split, achieving 99.98% Python syntax validity, 99.97% Flluma build success, and 99.14% STEP-export validity. For 9,909 predictions converted to point clouds, the mean normalized Chamfer Distance was 0.002124. This dataset supports various applications including conditioned CAD reconstruction, executable program synthesis, and editable reverse engineering.

Key takeaway

For AI Scientists and Machine Learning Engineers developing CAD generation or editing models, FllumaOne provides a uniquely validated, multimodal dataset. You can utilize its executable Python programs and detailed feature histories to train more robust models for tasks like conditioned CAD reconstruction or program synthesis. This resource significantly reduces data validation overhead, allowing you to focus on model innovation and achieve higher fidelity in your CAD applications.

Key insights

FllumaOne is a multimodal CAD dataset linking executable code to geometry and feature histories for advanced design research.

Principles

• Editable CAD research requires datasets exposing operations, parameters, and dependencies.
• Validated geometry and construction history are crucial for CAD datasets.
• Multimodal data improves CAD model understanding and synthesis.

Method

Models are generated by executable Python programs in Flluma, then validated via kernel geometry, solid validity, and export checks before inclusion in the dataset.

In practice

• Train models for conditioned CAD reconstruction.
• Develop systems for executable program synthesis.
• Explore B-Rep analysis and design completion.

Topics

• FllumaOne Dataset
• Parametric CAD
• Multimodal Datasets
• Program Synthesis
• Feature Trees
• B-Rep Analysis

Best for: Computer Vision Engineer, Research Scientist, AI Scientist, Machine Learning Engineer, AI Engineer

Related on AIssential

• DSAEval: Evaluating Data Science Agents on a Wide Range of Real-World Data Science Problems — Multimodal, multi-query benchmarks are crucial for evaluating LLM-based data science agents on real-world, open-ended problems.
• A multimodal large language model for materials science — MatterChat unifies material structure and text using a modular LLM, improving property prediction and scientific reasoning in materials science.
• 3DCodeBench: Benchmarking Agentic Procedural 3D Modeling Via Code — 3DCodeBench benchmarks VLM agents for procedural 3D modeling, revealing API mismatch issues and the need for better data and iterative refinement.
• Leveraging Multimodal LLMs for Built Environment and Housing Attribute Assessment from Street-View Imagery — Multimodal LLMs, fine-tuned and distilled, can automate large-scale building condition assessment from street-view imagery with high accuracy and efficiency.
• Agentic Large Language Models for Automated Structural Analysis of 3D Frame Systems — Agentic LLMs with structured decomposition and specialized agents can reliably automate complex 3D structural analysis tasks.
• JetBrains Releases Mellum2: A 12B MoE Model for Fast, Specialized Tasks in Multi-Model AI Pipelines — Mellum2 demonstrates a "focal model" approach, optimizing smaller MoE models for specific, fast tasks in AI pipelines.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.