Domain-specific models (ie, PubMedBERT) consistently outperformed the standard BERT model despite progressively decreasing training set sizes (P < .05).
確定! 回上一頁