arxiv.org
On the reliability of feature attribution methods for speech classification
As the capabilities of large-scale pre-trained models evolve, understanding the determinants of their outputs becomes more important. Feature attribution aims to reveal which parts of the input elemen...
