Transferable Adversarial Examples: Adversarial examples that are crafted for one model but can fool other models as well, it can be used to test the robustness of a machine learning model.