Self-supervised pre-training is essential for 3D point cloud representation learning, as annotating their irregular, topology-free structures is costly and labor-intensive. Masked autoencoders (MAEs) offer a promising framework but rely on explicit positional embeddings, such as patch center coordinates, which leak geometric information and limit data-driven structural learning. In this work, we propose Point-MaDi, a novel Point cloud Masked autoencoding Diffusion framework for pre-training that integrates a dual-diffusion pretext task into an MAE architecture to address this issue. Specifically, we introduce a center diffusion mechanism in the encoder, noising and predicting the coordinates of both visible and masked patch centers without ground-truth positional embeddings. These predicted centers are processed using a transformer with self-attention and cross-attention to capture intra- and inter-patch relationships. In the decoder, we design a conditional patch diffusion process, guided by the encoder's latent features and predicted centers to reconstruct masked patches directly from noise. This dual-diffusion design drives comprehensive global semantic and local geometric representations during pre-training, eliminating external geometric priors. Extensive experiments on ScanObjectNN, ModelNet40, ShapeNetPart, S3DIS, and ScanNet demonstrate that Point-MaDi achieves superior performance across downstream tasks, surpassing Point-MAE by 5.51\% on OBJ-BG, 5.17\% on OBJ-ONLY, and 4.34\% on PB-T50-RS for 3D object classification on the ScanObjectNN dataset.

Self-supervised Learning (SSL), including mainstream contrastive learning, has achieved significant success in learning visual representations without the need for data annotations in 3D vision. While most contrastive learning methods focus on instance-level information through random affine transformations, they pay limited attention to the intrinsic structures within point clouds. In this work, we propose a novel SSL paradigm for point cloud representation learning, called CCPoint, which incorporates a novel form of data corruption as a negative augmentation strategy. Specifically, we degrade the input point cloud with various corruptions and conduct contrastive learning among the augmented, raw, and corrupted points to learn robust and discriminative representations. To preserve the semantic structure of the point cloud even under heavy degradation, an auxiliary reconstruction decoder is introduced into the corruption branch to provide an additional supervision signal. We explore four families of corruptions—affine, noise, masking, and combined transformations. Different from previous methods that rely on multi-modal data or complex network architectures, CCPoint achieves state-of-the-art performance on three widely used datasets (ModelNet40, ScanObjectNN, and ShapeNetPart) with a lightweight and efficient structure, reaching top linear accuracies of 92.4% and 86.2% on ModelNet40 and ScanObjectNN, respectively.