Towards Reliable Objective Evaluation Metrics for Generative Singing Voice Separation Models

Overview of attention for article published in arXiv, October 2025

Altmetric Badge

About this Attention Score

Average Attention Score compared to outputs of the same age
Average Attention Score compared to outputs of the same age and source

Mentioned by

twitter: 2 X users

bluesky: 2 Bluesky users

Readers on

mendeley: 3 Mendeley

So far, Altmetric has seen 2 Bluesky posts from 2 Bluesky users, with an upper bound of 160 followers.

Degradation Category Rating listening test showed intrusive embedding-based metrics correlate better with DMOS than BSS-Eval for singing voice separation; MSE on MERT-L12 embeddings showed most balanced correlation across model types.

Paul A. Bereuter, Benjamin Stahl, Mark D. Plumbley, Alois Sontacchi: Towards Reliable Objective Evaluation Metrics for Generative Singing Voice Separation Models https://arxiv.org/abs/2507.11427 https://arxiv.org/pdf/2507.11427 https://arxiv.org/html/25…

This page shows the most recent Bluesky posts that mention this research output. Click here to find out how to access more activity.