In this work, we introduce a ``score-based assessment'' framework for estimating the transferability of pre-trained speech models (PSMs) for fine-tuning target tasks. We leverage upon two representation theories, Bayesian likelihood estimation and optimal transport, to generate rank scores for the PSM candidates using the extracted representations. Our framework efficiently computes transferability scores without actual fine-tuning of candidate models or layers by making a temporal independent hypothesis. We evaluate some popular supervised speech models (e.g., Conformer RNN-Transducer) and self-supervised speech models (e.g., HuBERT) in cross-layer and cross-model settings using public data. Experimental results show a high Spearman's rank correlation and low $p$-value between our estimation framework and fine-tuning ground truth. Our proposed transferability framework requires less computational time and resources, making it a resource-saving and time-efficient approach for tuning speech foundation models.

翻译：在本工作中，我们提出了一种基于评分的评估框架，用于估计预训练语音模型（PSM）在目标任务微调中的迁移性。我们利用贝叶斯似然估计和最优传输两种表示理论，通过提取的表示为候选PSM生成排名分数。通过引入时间独立假设，该框架无需实际微调候选模型或层，即可高效计算迁移性评分。我们使用公开数据在跨层和跨模型设置下评估了若干主流监督语音模型（如Conformer RNN-Transducer）和自监督语音模型（如HuBERT）。实验结果表明，我们的评估框架与微调真实值之间具有较高的斯皮尔曼等级相关系数及较低的p值。该迁移性评估框架计算时间和资源需求较低，为语音基础模型的调优提供了一种节省资源和时间的高效方法。