Self-supervised models trained for the study of the emergence of algebraic auditory structures.