PLLM: Pseudo-Labeling Large Language Models for CAD Program Synthesis

TL;DR

PLLM is a self-training framework that enables large language models to synthesize CAD programs from unlabeled 3D shapes by iteratively generating and refining candidate programs, improving fidelity and diversity.

Contribution

It introduces a novel self-training method for CAD program synthesis that does not require paired shape-program data, leveraging unlabeled datasets for fine-tuning LLMs.

Findings

Consistent improvements in geometric fidelity.

Enhanced program diversity.

Effective adaptation of CAD-Recode to unlabeled datasets.

Abstract

Recovering Computer-Aided Design (CAD) programs from 3D geometries is a widely studied problem. Recent advances in large language models (LLMs) have enabled progress in CAD program synthesis, but existing methods rely on supervised training with paired shape-program data, which is often unavailable. We introduce PLLM, a self-training framework for CAD program synthesis from unlabeled 3D shapes. Given a pre-trained CAD-capable LLM and a shape dataset, PLLM iteratively samples candidate programs, selects high-fidelity executions, and augments programs to construct synthetic program-shape pairs for fine-tuning. We experiment on adapting CAD-Recode from DeepCAD to the unlabeled ABC dataset show consistent improvements in geometric fidelity and program diversity.