arXiv:2610.41253v1 Announce Type: new Abstract: Point clouds have become a fundamental data format for emerging three dimensional (3D) applications, while their wireless delivery remains challenging due to the large data volume and irregular geometric structure. Existing semantic point cloud transmission methods are mainly developed for single-input single-output systems, making them difficult to apply directly to multiple-input multiple-output (MIMO) systems. To address this issue, we propose a partition-oriented semantic transmission framework for MIMO (POST-MIMO) point cloud delivery. POST-MIMO partitions the input point cloud into spatially coherent blocks and aligns them with parallel MIMO streams. Based on this block-aligned representation, the transmitter employs a hierarchical encoder to capture both intra-block local geometry and inter-block global dependency, while the auxiliary side information is transmitted jointly to assist reconstruction. At the receiver, a progressive decoder reconstructs the block geometry in a progressive manner and restores the final point cloud with block-aware adjustment. Experiments demonstrate improved reconstruction quality and lower computational resource consumption than representative reference schemes under the considered MIMO setting.