Robust Model Transfer

Hadi discusses how adding a robustness prior to models can lead to better transfer learning, despite lower accuracies in robust models. This approach shows promise in various tasks beyond image classification, like object detection and instance segmentation.