Robust Generative Audio Quality Assessment: Disentangling Quality from Spurious Correlations
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917349805785088 |
|---|---|
| author | Huang, Kuan-Tang Wang, Chien-Chun Yang, Cheng-Yeh Lee, Hung-Shin Wang, Hsin-Min Chen, Berlin |
| author_facet | Huang, Kuan-Tang Wang, Chien-Chun Yang, Cheng-Yeh Lee, Hung-Shin Wang, Hsin-Min Chen, Berlin |
| contents | The rapid proliferation of AI-Generated Content (AIGC) has necessitated robust metrics for perceptual quality assessment. However, automatic Mean Opinion Score (MOS) prediction models are often compromised by data scarcity, predisposing them to learn spurious correlations-- such as dataset-specific acoustic signatures-- rather than generalized quality features. To address this, we leverage domain adversarial training (DAT) to disentangle true quality perception from these nuisance factors. Unlike prior works that rely on static domain priors, we systematically investigate domain definition strategies ranging from explicit metadata-driven labels to implicit data-driven clusters. Our findings reveal that there is no "one-size-fits-all" domain definition; instead, the optimal strategy is highly dependent on the specific MOS aspect being evaluated. Experimental results demonstrate that our aspect-specific domain strategy effectively mitigates acoustic biases, significantly improving correlation with human ratings and achieving superior generalization on unseen generative scenarios. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_16201 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Robust Generative Audio Quality Assessment: Disentangling Quality from Spurious Correlations Huang, Kuan-Tang Wang, Chien-Chun Yang, Cheng-Yeh Lee, Hung-Shin Wang, Hsin-Min Chen, Berlin Audio and Speech Processing Artificial Intelligence Sound Signal Processing The rapid proliferation of AI-Generated Content (AIGC) has necessitated robust metrics for perceptual quality assessment. However, automatic Mean Opinion Score (MOS) prediction models are often compromised by data scarcity, predisposing them to learn spurious correlations-- such as dataset-specific acoustic signatures-- rather than generalized quality features. To address this, we leverage domain adversarial training (DAT) to disentangle true quality perception from these nuisance factors. Unlike prior works that rely on static domain priors, we systematically investigate domain definition strategies ranging from explicit metadata-driven labels to implicit data-driven clusters. Our findings reveal that there is no "one-size-fits-all" domain definition; instead, the optimal strategy is highly dependent on the specific MOS aspect being evaluated. Experimental results demonstrate that our aspect-specific domain strategy effectively mitigates acoustic biases, significantly improving correlation with human ratings and achieving superior generalization on unseen generative scenarios. |
| title | Robust Generative Audio Quality Assessment: Disentangling Quality from Spurious Correlations |
| topic | Audio and Speech Processing Artificial Intelligence Sound Signal Processing |
| url | https://arxiv.org/abs/2603.16201 |