MADB

dataset·music aesthetics·active

Large music-aesthetics benchmark with multi-dimensional professional ratings and textual comments.

Recorded facts

Official site	https://github.com/knownree/madb ↗
Geography	Global
arxiv	2607.06929
github repository	knownree/madb
size	tracks: 9999 · trained annotators: 30 · perceptual dimensions: 10 · ratings per track approx: 10
scope	Human-aligned music aesthetic assessment and multimodal analysis.
genres	not publicly normalized
steward	MADB authors
creators	Sirui Zhang; Tianle Wang; Xinyi Tong; Peiyang Yu; Jishang Chen; Liangke Zhao; Haoxin Zhang; Duo Xu; Xin Jin; Feng Yu; Songchun Zhu
languages	not publicly normalized
modalities	music tracks; 10 perceptual ratings; overall score; textual comments
geographies	Global
access method	Project GitHub referenced by paper.
intended uses	music aesthetic assessment; human-aligned evaluation
related papers	MADB: A Large-Scale Music Aesthetics Dataset
annotation method	Thirty trained annotators; about ten ratings per track across ten dimensions and an overall score.
collection method	Tracks selected for large-scale professional aesthetic annotation.
current availability	Project page exists; release completeness requires review.
us market scope basis	Publicly available or materially used in US-facing music/audio-AI research.
limitations biases disputes	Very recent preprint with limited independent use and replication.; Track sources, genres and annotator demographics need audit.