AFMG dedicated to advancing general-purpose Audio Foundation Models that can understand, generate, and reason over speech, music, and sound