MOS-Bench
MOS‑Bench is a collection of datasets for training and evaluating the generalization ability of subjective speech quality assessment (SSQA) models, developed by Nagoya University. The collection comprises seven training sets and twelve test sets, covering various sampling rates, languages, and speech types, including synthetic speech generated by text‑to‑speech (TTS), voice conversion (VC), and speech enhancement (SE) systems, as well as non‑synthetic speech such as transmission, noise, and reverberated speech. The dataset was created by integrating and processing multiple listening test datasets, aiming to address the generalization challenges of speech quality assessment models on unseen data. MOS‑Bench is widely used in speech processing, especially for research on subjective speech quality evaluation.