Mesh-Pro：面向艺术家风格四边形网格生成的异步优势引导排序偏好优化
Mesh-Pro: Asynchronous Advantage-guided Ranking Preference Optimization for Artist-style Quadrilateral Mesh Generation
提出异步在线强化学习框架和 ARPO 算法，可直接用于提升 3D 网格生成的后训练效率与泛化能力；对角感知的混合三角-四边形 tokenization 和基于射线的奖励为几何完整性提供了新思路。
原文依据 we design the first asynchronous online RL framework tailored for 3D mesh generation post-training efficiency improvement
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论强化学习（RL）在文本和图像生成中已取得显著成功，但其在 3D 生成中的潜力仍 largely 未被探索。现有尝试通常依赖离线直接偏好优化（DPO）方法，存在训练效率低和泛化能力有限的问题。在本工作中，我们旨在提升 RL 在 3D 网格生成中的训练效率和生成质量。具体而言：（1）我们设计了首个专为提升 3D 网格生成后训练效率而定制的异步在线 RL 框架，比同步 RL 快 3.75 倍。（2）我们提出了优势引导排序偏好优化（ARPO），这是一种新颖的 RL 算法，在训练效率和泛化能力之间取得了比当前为 3D 网格生成设计的 RL 算法（如 DPO 和组相对策略优化（GRPO））更好的权衡。（3）基于异步 ARPO，我们提出了 Mesh-Pro，额外引入了一种新颖的对角感知混合三角-四边形 tokenization 用于网格表示，以及一种基于射线的奖励用于几何完整性。Mesh-Pro 在艺术性和密集网格上达到了最优性能。
英文原文摘要 1203 字符
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 3.75xfaster 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.