PointNSP：基于下一尺度细节层次预测的自回归三维点云生成
PointNSP: Autoregressive 3D Point Cloud Generation with Next-Scale Level-of-Detail Prediction
提出基于下一尺度预测的自回归点云生成框架，利用多尺度分解对齐点集的无序性，避免固定顺序偏差，可复用于其他无序集合生成任务（如点云补全、场景生成）。
原文依据 we propose PointNSP, a coarse-to-fine generative framework that preserves global shape structure at low resolutions and progressively refines fine-grained geometry at higher scales through a next-scale prediction paradigm.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论自回归点云生成在质量上长期落后于基于扩散的方法。这一性能差距源于自回归模型对本质上无序的点集施加了人为的顺序，迫使形状生成以一系列局部预测的形式进行。这种顺序偏差强化了短程连续性，但限制了模型捕捉长程依赖的能力，从而削弱了其施加全局结构属性（如对称性、几何一致性和大尺度空间规律性）的能力。受形状建模中细节层次（LoD）原理的启发，我们提出了 PointNSP，一个由粗到细的生成框架，该框架在低分辨率下保留全局形状结构，并通过下一尺度预测范式在高尺度上逐步细化细粒度几何。这种多尺度分解使自回归目标与点集的置换不变性对齐，在避免脆弱的固定顺序的同时实现了丰富的尺度内交互。严格遵循基线实验设置，在 ShapeNet 基准上的实证结果表明，PointNSP 首次在自回归范式内实现了最优的生成质量。此外，它在参数、训练和推理效率上超越了强大的基于扩散的基线。最后，在 8,192 个点的密集生成下，PointNSP 的优势更加明显，凸显了其强大的可扩展潜力。
英文原文摘要 1546 字符
Autoregressive point cloud generation has long lagged behind diffusion-based approaches in quality. The performance gap stems from the fact that autoregressive models impose an artificial ordering on inherently unordered point sets, forcing shape generation to proceed as a sequence of local predictions. This sequential bias reinforces short-range continuity but limits the model's ability to capture long-range dependencies, thereby weakening its capacity to enforce global structural properties such as symmetry, geometric consistency, and large-scale spatial regularities. Inspired by the level-of-detail (LoD) principle in shape modeling, we propose PointNSP, a coarse-to-fine generative framework that preserves global shape structure at low resolutions and progressively refines fine-grained geometry at higher scales through a next-scale prediction paradigm. This multi-scale factorization aligns the autoregressive objective with the permutation-invariant nature of point sets, enabling rich intra-scale interactions while avoiding brittle fixed orderings. Strictly following the baseline experimental setups, empirical results on ShapeNet benchmark demonstrate that PointNSP achieves state-of-the-art (SoTA) generation quality for the first time within the autoregressive paradigm. Moreover, it surpasses strong diffusion-based baselines in parameter, training, and inference efficiency. Finally, under dense generation with 8,192 points, PointNSP's advantages become even more pronounced, highlighting its strong scalability potential.