Teaching a Machine to Hear Music Like You Do

Can a machine separate melody, rhythm, and timbre the way your ear does? MERIT trains three lightweight modules on one model. Each hits over 99.6% accuracy on its own dimension while staying near random chance on the others. Clean separation, finally.