| title | MERIT: Learning Disentangled Music Representations for Audio Similarity |
|---|---|
| authors | Abhinaba Roy, Junyi Liang, Dorien Herremans |
| date | 2026-05-26 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2605.27346 |
| affiliation | AMAAI Lab |
| upvotes | 5 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | Current music similarity models typically compute a single, monolithic score, entangling distinct musical dimensions like melody, rhythm, and timbre. This limits user control and interpretability, making it impossible to execute nuanced queries. We introduce MERIT, a framework for learning disentangled, factor-specific music representations tailored to these three core dimensions. To overcome the lack of isolated musical variations in real-world audio, we use a novel training strategy that uses conditional audio generation and source-separated stems to strongly encourage single-factor variation in training data. Our evaluations demonstrate strong factor-wise disentanglement. Each head responds strongly to its intended perceptual dimension while remaining near chance on the others, a representational property that holds across both the synthetic training domain and independent real-world audio. |
cite
@misc{roy2026merit,
title={MERIT: Learning Disentangled Music Representations for Audio Similarity},
author={Abhinaba Roy and Junyi Liang and Dorien Herremans},
year={2026},
eprint={2605.27346},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2605.27346}
}