1 paper · 1 filter
Kyohoon Jin, Juhwan Choi, Jungmin Yun +3
Deep learning models often learn and exploit spurious correlations in training data, using these non-target features to inform their predictions. Such reliance leads to performance…