MMAU

dataset·general audio reasoning·active

A multi-domain benchmark of expert audio understanding and reasoning spanning speech, environmental sound and music.

Recorded facts

Official site	https://sakshi113.github.io/mmau_homepage/ ↗
Geography	United States
size	skills: 27 · questions: 10000 · source corpora: 13
scope	Information extraction and complex reasoning across speech, sound and music.
genres	multiple
license	mixed/source-specific
steward	University of Maryland-led collaboration
creators	MMAU authors
languages	primarily English questions; source audio varies
modalities	audio; multiple-choice questions; answers; skill taxonomy
geographies	Global
access method	Official project links code and data.
intended uses	audio-language model evaluation; audio reasoning evaluation
consent claims	Inherited from source corpora; annotator terms not verified.
related papers	MMAU: A Massive Multi-Task Audio Understanding and Reasoning Benchmark
annotation method	Expert-oriented question design with answer choices and skill labels.
collection method	Questions and audio are assembled across thirteen source corpora.
provenance claims	Official page documents source domains and benchmark organization.
current availability	Project page and artifacts available.
commercial use limits	Depends on each source corpus.
us market scope basis	Publicly available or materially used in US-facing music/audio-AI research.
limitations biases disputes	Multiple-choice accuracy is not equivalent to open-ended auditory competence.; Music is only one of three domains.