pipette
ESEspañol

CloudEar: Integrating Perceptual, Diagnostic, and Cross-Modal Evidence for Music Evaluation

Qiqi He, Anqi Huang

Preprint

In the authors' words

Evaluating generated music requires modeling perceptual quality, technical audio quality, and prompt alignment. Existing methods often overlook signal-level defects, produce compressed quality scores, and miss fine-grained musical attributes in text prompts. We propose CloudEar, a pairwise framework for evaluating musicality and prompt alignment. It comprises three components: i) a Perceptual Expert for song-quality assessment; ii) a Diagnostic Expert for learned and signal-level quality analysis; and iii) a Cross-modal Expert for prompt-song alignment based on musical attributes. Experiments show that CloudEar outperforms the compared baselines in pairwise preference accuracy and score correlation.

Main resultThe abstract does not state a limitation.

Appeared: Friday, October 9. arXiv. Preprint, not yet peer-reviewed.

Authors' comment: Accepted at the Late-Breaking/Demo session of ISMIR 2026