PointNSP: Autoregressive 3D Point Cloud Generation with Next-Scale Level-of-Detail Prediction

Ziqiao Meng1 Qichao Wang2 Zhiyang Dou3 Zixing Song4 Zhipeng Zhou2
Irwin King5 Peilin Zhao6
1National University of Singapore 2Nanyang Technological University
3University of Hong Kong 4University of Cambridge
5The Chinese University of Hong Kong 6Shanghai Jiao Tong University

[Paper] [Code]

paradigm figure

Three types of point cloud generative models: (a) diffusion-based methods that iteratively denoise shapes starting from Gaussian noise; (b) vanilla autoregressive (AR) methods that predict the next point by flattening the 3D shape into a sequence; and (c) our proposed PointNSP, which predicts next-scale level-of-detail in a coarse-to-fine manner.

paradigm figure

Overview of PoinNSP: Illustration of training a multi-scale VQ-VAE for 3D point cloud reconstruction across scales \( s_{1} \) to \( s_{4} \), resulting in a multi-scale token sequence \( Q = (q_{1}, \dots, q_{4}) \).

paradigm figure

Illustration of training a multi-scale 3D point cloud causal transformer with intermediate decoding, upsampling, position-aware soft masks, and block-wise causal masks.

paradigm figure

Our generation results (right) compared to baseline models (left). PointNSP generates high-quality and diverse 3D point clouds.

paradigm figure

Visualization of multi-scale point clouds during the PointNSP generation process as the scale K increases.

completion figure

Visualizations of point cloud completion results.

upsampling figure

Visualizations of point cloud upsampling results.

Citation

To cite the paper, please use the below:

@misc{meng2025pointnspautoregressive3dpoint,
    title={PointNSP: Autoregressive 3D Point Cloud Generation with Next-Scale Level-of-Detail Prediction}, 
    author={Ziqiao Meng and Qichao Wang and Zhiyang Dou and Zixing Song and Zhipeng Zhou and Irwin King and Peilin Zhao},
    year={2025},
    eprint={2510.05613},
    archivePrefix={arXiv},
    primaryClass={cs.CV},
    url={https://arxiv.org/abs/2510.05613}, 
}