Concept-based attacks target neural-network logic, not class labels
Retraining on the attacks kept classical robustness gains while better removing semantic shortcuts.
Stories per day · 30 days
1 stories, newest first · one card per story
1 stories, newest first · one card per story
Retraining on the attacks kept classical robustness gains while better removing semantic shortcuts.
Today 1.5K readers · 2.4K views · 768 reads · all time 189K views