Recently, multiple models proposed for link prediction (LP) demonstrate
impressive results on benchmark datasets. However, many popular benchmark
datasets often assume that dataset samples are drawn from the same distribution
(i.e., IID samples). In real-world situations, this assumpti