Research on evaluating and labeling AI audio: AI judges vs human listeners, evaluation metrics, benchmarks, listening tests, and how to source and annotate audio training data.