ALMs are multimodal models that take interleaved audio and text as input and output text, powering use cases from ...