Learning an Invertible Output Mapping Can Mitigate Simplicity Bias in Neural Networks

AAML
Re-assign community