MMSU

An audio understanding benchmark that evaluates a model's ability to select the correct answer from multiple options, assessing comprehension of audio content across varied tasks through multiple-choice evaluation.