AudioSAE: Towards Understanding of Audio-Processing Models with Sparse AutoEncoders - Explained Simply | ArXiv Explained