This line of research focuses on relational graph learning, where nodes and edges carry rich relational information. We introduce methods that combine Gaussian modeling with graph transformers to capture uncertainty and complex relational patterns in graph data.