Mesh-Pro: Asynchronous Advantage-guided Ranking Preference Optimization for Artist-style Quadrilateral Mesh Generation

2026-02-28

AI Summary

This work addresses the inefficiency and limited generalization of existing offline preference optimization methods in 3D mesh generation by proposing the first asynchronous online reinforcement learning framework tailored for this task. The framework introduces Advantage-guided Ranked Preference Optimization (ARPO), a diagonal-aware hybrid triangle–quadrilateral tokenization representation, and a geometry-completeness reward mechanism based on ray sampling. Experimental results demonstrate that the proposed method achieves a 3.75× speedup in training compared to synchronous reinforcement learning and attains state-of-the-art performance in generating artistically styled and high-density quadrilateral meshes.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Learning Preferences or Rankings
• Knowledge Representation and Reasoning: Preferences

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Reinforcement learning (RL) has demonstrated remarkable success in text and image generation, yet its potential in 3D generation remains largely unexplored. Existing attempts typically rely on offline direct preference optimization (DPO) method, which suffers from low training efficiency and limited generalization. In this work, we aim to enhance both the training efficiency and generation quality of RL in 3D mesh generation. Specifically, (1) we design the first asynchronous online RL framework tailored for 3D mesh generation post-training efficiency improvement, which is 3.75$\times$ faster than synchronous RL. (2) We propose Advantage-guided Ranking Preference Optimization (ARPO), a novel RL algorithm that achieves a better trade-off between training efficiency and generalization than current RL algorithms designed for 3D mesh generation, such as DPO and group relative policy optimization (GRPO). (3) Based on asynchronous ARPO, we propose Mesh-Pro, which additionally introduces a novel diagonal-aware mixed triangular-quadrilateral tokenization for mesh representation and a ray-based reward for geometric integrity. Mesh-Pro achieves state-of-the-art performance on artistic and dense meshes.

Problem

• 3D mesh generation
• reinforcement learning
• training efficiency
• generalization
• preference optimization

Innovation

• asynchronous reinforcement learning
• Advantage-guided Ranking Preference Optimization
• quadrilateral mesh generation
• diagonal-aware tokenization
• ray-based reward