MMAR: A Challenging Benchmark for Deep Reasoning in Speech, Audio, Music, and Their Mix | Read Paper on Bytez