The paper introduces Point Mamba, a novel point cloud processing backbone based on the state space model (SSM). The authors address the challenge of extending SSM to point clouds, which are inherently disordered and irregular, by proposing an octree-based ordering mechanism to establish causality. This mechanism globally sorts points in a z-order sequence while retaining their spatial proximity. The Point Mamba architecture includes an octree establishment layer, a feature embedding layer, and a series of Point Mamba blocks and downsampling layers. The method achieves state-of-the-art performance on the ModelNet40 classification dataset with 93.4% accuracy and on the ScanNet semantic segmentation dataset with 75.7 mIOU, outperforming transformer-based methods. Point Mamba also demonstrates linear complexity, making it more efficient than transformer-based approaches. The paper discusses the contributions, related work, method details, experiments, and future directions, highlighting the potential of SSM as a generic backbone in point cloud understanding.The paper introduces Point Mamba, a novel point cloud processing backbone based on the state space model (SSM). The authors address the challenge of extending SSM to point clouds, which are inherently disordered and irregular, by proposing an octree-based ordering mechanism to establish causality. This mechanism globally sorts points in a z-order sequence while retaining their spatial proximity. The Point Mamba architecture includes an octree establishment layer, a feature embedding layer, and a series of Point Mamba blocks and downsampling layers. The method achieves state-of-the-art performance on the ModelNet40 classification dataset with 93.4% accuracy and on the ScanNet semantic segmentation dataset with 75.7 mIOU, outperforming transformer-based methods. Point Mamba also demonstrates linear complexity, making it more efficient than transformer-based approaches. The paper discusses the contributions, related work, method details, experiments, and future directions, highlighting the potential of SSM as a generic backbone in point cloud understanding.