CAL-MOS: Bridging Layers with Adapters for Robust MOS Prediction Across Speech Foundation Models

2026-09-13

AI Summary

研究解决了多层信息融合在语音质量评估中的不稳定问题，通过使用层适配器方法提高了跨模型和数据集的鲁棒性。

Abstract

Speech Quality Assessment (SQA) is essential for modern speech technologies, and recent non-intrusive SQA predictors increasingly rely on Speech Foundation Models (SFMs). However, because SFMs expose representations from many layers, it remains unclear which depths are most informative for MOS prediction and how multi-layer information should be combined reliably across backbones and datasets. We benchmark ten SFMs on four MOS datasets under three regimes: full fine-tuning, last-layer probing with a frozen encoder, and naive cross-layer weighted aggregation. We find that the best layer is strongly backbone- and dataset-dependent, and that naive weighted fusion can be unstable across settings. We further evaluate a layer-calibrated aggregation variant that applies per-layer adapters before pooling, which improves the robustness of multi-layer fusion and narrows the gap to full fine-tuning while keeping the backbone frozen.

Problem

• Speech Quality Assessment
• Speech Foundation Models
• MOS prediction
• layer representation
• multi-layer information

Innovation

• Layer-calibrated Aggregation
• Adapters
• Speech Foundation Models (SFMs)
• Multi-layer Fusion
• Robust MOS Prediction