LLM-to-Phy3D: Physically Conform Online 3D Object Generation with LLMs

TL;DR

LLM-to-Phy3D enhances existing LLM-based 3D object generation by incorporating physical and visual evaluations, enabling the creation of physically conforming and innovative 3D designs through an iterative refinement process.

Contribution

This paper introduces a novel online refinement loop that integrates physical and visual feedback to improve LLM-to-3D models for physically viable and innovative object generation.

Findings

Achieved 4.5% to 106.7% improvements in physical conformity of 3D designs.

Demonstrated effectiveness in vehicle design optimization.

Validated potential for broader applications in Physical AI.

Abstract

The emergence of generative artificial intelligence (GenAI) and large language models (LLMs) has revolutionized the landscape of digital content creation in different modalities. However, its potential use in Physical AI for engineering design, where the production of physically viable artifacts is paramount, remains vastly underexplored. The absence of physical knowledge in existing LLM-to-3D models often results in outputs detached from real-world physical constraints. To address this gap, we introduce LLM-to-Phy3D, a physically conform online 3D object generation that enables existing LLM-to-3D models to produce physically conforming 3D objects on the fly. LLM-to-Phy3D introduces a novel online black-box refinement loop that empowers large language models (LLMs) through synergistic visual and physics-based evaluations. By delivering directional feedback in an iterative refinement process, LLM-to-Phy3D actively drives the discovery of prompts that yield 3D artifacts with enhanced physical performance and greater geometric novelty relative to reference objects, marking a substantial contribution to AI-driven generative design. Systematic evaluations of LLM-to-Phy3D, supported by ablation studies in vehicle design optimization, reveal various LLM improvements gained by 4.5% to 106.7% in producing physically conform target domain 3D designs over conventional LLM-to-3D models. The encouraging results suggest the potential general use of LLM-to-Phy3D in Physical AI for scientific and engineering applications.