QuadGPT: Native Quadrilateral Mesh Generation with Autoregressive Models

2025-09-25

AI Summary

Existing quad-mesh generation methods predominantly rely on a two-stage “triangle mesh → quad merging” paradigm, resulting in poor topological quality and severe geometric distortion. This paper proposes the first end-to-end autoregressive quad-mesh generation framework. First, we design a native quad-aware unified hybrid-topology tokenization scheme, enabling sequence-based modeling of meshes with arbitrary connectivity. Second, we introduce a topology-aware reinforcement learning fine-tuning strategy—topology-directed Direct Preference Optimization (tDPO)—that explicitly optimizes quad quality, regularity, and connectivity metrics. Our approach eliminates dependence on triangular base meshes and directly synthesizes high-quality, quad-dominant meshes. Experiments demonstrate significant improvements over state-of-the-art triangle-to-quad methods: 32% reduction in Chamfer distance (geometric accuracy), 18% increase in quad ratio, and 41% reduction in singularity count (topological quality). This work establishes a novel native quad-generation paradigm.

Technology Category

• Natural Language Processing: Generation
• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

The generation of quadrilateral-dominant meshes is a cornerstone of professional 3D content creation. However, existing generative models generate quad meshes by first generating triangle meshes and then merging triangles into quadrilaterals with some specific rules, which typically produces quad meshes with poor topology. In this paper, we introduce QuadGPT, the first autoregressive framework for generating quadrilateral meshes in an end-to-end manner. QuadGPT formulates this as a sequence prediction paradigm, distinguished by two key innovations: a unified tokenization method to handle mixed topologies of triangles and quadrilaterals, and a specialized Reinforcement Learning fine-tuning method tDPO for better generation quality. Extensive experiments demonstrate that QuadGPT significantly surpasses previous triangle-to-quad conversion pipelines in both geometric accuracy and topological quality. Our work establishes a new benchmark for native quad-mesh generation and showcases the power of combining large-scale autoregressive models with topology-aware RL refinement for creating structured 3D assets.

Problem

• Generates quadrilateral meshes end-to-end without triangle conversion
• Improves geometric accuracy and topological quality of quad meshes
• Handles mixed triangle-quadrilateral topologies with reinforcement learning refinement

Innovation

• Autoregressive framework generates quadrilateral meshes end-to-end
• Unified tokenization handles mixed triangle and quadrilateral topologies
• Specialized RL fine-tuning enhances generation quality