PR-CAD: Progressive Refinement for Unified Controllable and Faithful Text-to-CAD Generation with Large Language Models
Summary
PR-CAD introduces a progressive refinement framework for unified text-to-CAD generation with LLMs, combining generation and editing into an all-in-one system. It includes a CAD-focused interaction dataset, a tailored representation for LLMs, and a reinforcement learning–driven reasoning module, achieving state-of-the-art controllability and faithfulness in CAD modeling.