Select a category to explore research frontiers
Loading categories...
Research examines defences that reduce model susceptibility to deliberately crafted inputs. Robustness methods generally trade some ordinary recognition accuracy for attack resistance.