CAD-Coder: Text-to-CAD Generation with Chain-of-Thought and Geometric Reward

TL;DR

CAD-Coder is a new framework that converts text descriptions into parametric CAD scripts, enabling geometric validation and improved reasoning through a two-stage learning process and chain-of-thought planning.

Contribution

It introduces a novel text-to-CAD generation method using CadQuery scripts, a two-stage training pipeline with reinforcement learning, and a large dataset for training and evaluation.

Findings

Enables LLMs to generate diverse, valid CAD models from natural language.

Improves geometric fidelity using Chamfer Distance-based rewards.

Achieves state-of-the-art performance in text-to-CAD generation.

Abstract

In this work, we introduce CAD-Coder, a novel framework that reformulates text-to-CAD as the generation of CadQuery scripts - a Python-based, parametric CAD language. This representation enables direct geometric validation, a richer modeling vocabulary, and seamless integration with existing LLMs. To further enhance code validity and geometric fidelity, we propose a two-stage learning pipeline: (1) supervised fine-tuning on paired text-CadQuery data, and (2) reinforcement learning with Group Reward Policy Optimization (GRPO), guided by a CAD-specific reward comprising both a geometric reward (Chamfer Distance) and a format reward. We also introduce a chain-of-thought (CoT) planning process to improve model reasoning, and construct a large-scale, high-quality dataset of 110K text-CadQuery-3D model triplets and 1.5K CoT samples via an automated pipeline. Extensive experiments demonstrate that CAD-Coder enables LLMs to generate diverse, valid, and complex CAD models directly from natural language, advancing the state of the art of text-to-CAD generation and geometric reasoning.