Rho-Perfect: Correlation Ceiling For Subjective Evaluation Datasets

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autore principale: Cumlin, Fredrik
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914315638931456
author Cumlin, Fredrik
author_facet Cumlin, Fredrik
contents Subjective ratings contain inherent noise that limits the model-human correlation, but this reliability issue is rarely quantified. In this paper, we present $ρ$-Perfect, a practical estimation of the highest achievable correlation of a model on subjectively rated datasets. We define $ρ$-Perfect to be the correlation between a perfect predictor and human ratings, and derive an estimate of the value based on heteroscedastic noise scenarios, a common occurrence in subjectively rated datasets. We show that $ρ$-Perfect squared estimates test-retest correlation and use this to validate the estimate. We demonstrate the use of $ρ$-Perfect on a speech quality dataset and show how the measure can distinguish between model limitations and data quality issues.
format Preprint
id arxiv_https___arxiv_org_abs_2602_08552
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Rho-Perfect: Correlation Ceiling For Subjective Evaluation Datasets
Cumlin, Fredrik
Machine Learning
Audio and Speech Processing
Subjective ratings contain inherent noise that limits the model-human correlation, but this reliability issue is rarely quantified. In this paper, we present $ρ$-Perfect, a practical estimation of the highest achievable correlation of a model on subjectively rated datasets. We define $ρ$-Perfect to be the correlation between a perfect predictor and human ratings, and derive an estimate of the value based on heteroscedastic noise scenarios, a common occurrence in subjectively rated datasets. We show that $ρ$-Perfect squared estimates test-retest correlation and use this to validate the estimate. We demonstrate the use of $ρ$-Perfect on a speech quality dataset and show how the measure can distinguish between model limitations and data quality issues.
title Rho-Perfect: Correlation Ceiling For Subjective Evaluation Datasets
topic Machine Learning
Audio and Speech Processing
url https://arxiv.org/abs/2602.08552