Abstract: Out-of-distribution (OOD) detection has garnered increasing concern for identifying test samples that exhibit a distributional shift from the training dataset in practical deep learning ...