Zero-to-CAD: Agentic Synthesis of Interpretable CAD Programs at Million-Scale Without Real Data

2026-04-27

AI Summary

Existing large-scale 3D datasets lack parametric modeling histories that capture design intent, hindering the learning of editable and interpretable CAD programs. This work proposes an agent-based, feedback-driven framework that integrates a large language model into a CAD environment to iteratively generate, execute, and verify code, enabling, for the first time, large-scale synthesis of interpretable CAD programs without requiring ground-truth modeling histories. By combining tool invocation, documentation retrieval, and geometric validation, the method supports diverse modeling operations and successfully synthesizes approximately one million executable, human-readable, and editable CAD programs, with a high-quality subset of 100,000 programs publicly released. A vision-language model fine-tuned on this dataset significantly outperforms strong baselines such as GPT-5.2 on the task of reconstructing CAD programs from multi-view images.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Computer Vision: Large Vision Models
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Large language models for search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Computer-Aided Design (CAD) models are defined by their construction history: a parametric recipe that encodes design intent. However, existing large-scale 3D datasets predominantly consist of boundary representations (B-Reps) or meshes, stripping away this critical procedural information. To address this scarcity, we introduce Zero-to-CAD, a scalable framework for synthesizing executable CAD construction sequences. We frame synthesis as an agentic search problem: by embedding a large language model (LLM) within a feedback-driven CAD environment, our system iteratively generates, executes, and validates code using tools and documentation lookup to promote geometric validity and operation diversity. This agentic approach enables the synthesis of approximately one million executable, readable, editable CAD sequences, covering a rich vocabulary of operations beyond sketch-and-extrude workflows. We also release a curated subset of 100,000 high-quality models selected for geometric diversity. To demonstrate the dataset's utility, we fine-tune a vision-language model on our synthetic data to reconstruct editable CAD programs from multi-view images, outperforming strong baselines, including GPT-5.2, and effectively bootstrapping sequence generation capabilities without real construction-history training data. Zero-to-CAD bridges the gap between geometric scale and parametric interpretability, offering a vital resource for the next generation of CAD AI.

Problem

• CAD synthesis
• construction history
• parametric modeling
• procedural representation
• 3D dataset scarcity

Innovation

• Agentic CAD synthesis
• Executable CAD programs
• Large language models (LLMs)
• Parametric modeling
• Synthetic data generation