Abstract: Recent graph neural network-based collaborative filtering models are capable of representing complex user-item interaction information, which is crucial when datasets exhibit high sparsity.