In the domain of heterogeneous graph representation learning, traditional methodologies often rely excessively on manually crafted meta-paths and neighbor aggregation mechanisms when faced with ...