**HYPEBOY: Generative Self-Supervised Representation Learning on Hypergraphs** This paper introduces HYPEBOY, a novel generative self-supervised learning (SSL) method for hypergraphs. Hypergraphs are complex structures that capture higher-order interactions among multiple nodes through hyperedges. Effective representation learning on hypergraphs is crucial for tasks like node classification and hyperedge prediction. HYPEBOY addresses the challenge of designing a generative SSL strategy for hypergraphs by formulating a generative SSL task called hyperedge filling. This task involves predicting a missing node in a hyperedge based on the other nodes in the hyperedge. The method is designed to learn effective general-purpose hypergraph representations, outperforming 16 baseline methods across 11 benchmark datasets. HYPEBOY is based on the hyperedge filling task, which is theoretically connected to node classification. The method includes a two-stage training scheme to enhance performance. The first stage involves feature and topology augmentation to mitigate over-reliance on proximity information. The second stage involves hyperedge filling loss computation, which is designed to ensure alignment and uniformity of node representations. Theoretical analysis shows that the hyperedge filling task improves node classification accuracy by refining representations to better capture the hypergraph structure. Experiments demonstrate that HYPEBOY outperforms existing SSL methods in both node classification and hyperedge prediction tasks. The method achieves the best average ranking across multiple datasets, indicating its effectiveness in learning general-purpose hypergraph representations. The results highlight the importance of preserving higher-order interactions in hypergraph representation learning and validate the design choices of HYPEBOY. The code and datasets are available at https://github.com/kswoo97/hypeboy.**HYPEBOY: Generative Self-Supervised Representation Learning on Hypergraphs** This paper introduces HYPEBOY, a novel generative self-supervised learning (SSL) method for hypergraphs. Hypergraphs are complex structures that capture higher-order interactions among multiple nodes through hyperedges. Effective representation learning on hypergraphs is crucial for tasks like node classification and hyperedge prediction. HYPEBOY addresses the challenge of designing a generative SSL strategy for hypergraphs by formulating a generative SSL task called hyperedge filling. This task involves predicting a missing node in a hyperedge based on the other nodes in the hyperedge. The method is designed to learn effective general-purpose hypergraph representations, outperforming 16 baseline methods across 11 benchmark datasets. HYPEBOY is based on the hyperedge filling task, which is theoretically connected to node classification. The method includes a two-stage training scheme to enhance performance. The first stage involves feature and topology augmentation to mitigate over-reliance on proximity information. The second stage involves hyperedge filling loss computation, which is designed to ensure alignment and uniformity of node representations. Theoretical analysis shows that the hyperedge filling task improves node classification accuracy by refining representations to better capture the hypergraph structure. Experiments demonstrate that HYPEBOY outperforms existing SSL methods in both node classification and hyperedge prediction tasks. The method achieves the best average ranking across multiple datasets, indicating its effectiveness in learning general-purpose hypergraph representations. The results highlight the importance of preserving higher-order interactions in hypergraph representation learning and validate the design choices of HYPEBOY. The code and datasets are available at https://github.com/kswoo97/hypeboy.