MIT Press

Table 3:

Spearman correlation ρ for the STSb dataset under the supervised setting. We use *-NLI to denote the model additionally trained on NLI datasets. ♯ indicates that results are reproduced by ourselves; § indicates results are taken from Reimers and Gurevych (2019); Surrogate are results for our proposed method.

Model	Spearman ρ
BERT [CLS]^sharp	73.01
BERT_base^§	84.30
SBERT_base^§	84.67
SRoBERTa_base^§	84.92
Surrogate_base	87.91

BERT-NLI_base^§	88.33
SBERT-NLI_base^§	85.35
SRoBERTa-NLI_base^§	84.79
Surrogate-NLI_base	89.95

BERT_large^§	85.64
SBERT_large^§	84.45
SRoBERTa_large^§	85.02
Surrogate_large	88.52

BERT-NLI_large^§	88.77
SBERT-NLI_large^§	86.10
SRoBERTa-NLI_large^§	86.15
Surrogate-NLI_large	90.69

Model	Spearman ρ
BERT [CLS]^sharp	73.01
BERT_base^§	84.30
SBERT_base^§	84.67
SRoBERTa_base^§	84.92
Surrogate_base	87.91

BERT-NLI_base^§	88.33
SBERT-NLI_base^§	85.35
SRoBERTa-NLI_base^§	84.79
Surrogate-NLI_base	89.95

BERT_large^§	85.64
SBERT_large^§	84.45
SRoBERTa_large^§	85.02
Surrogate_large	88.52

BERT-NLI_large^§	88.77
SBERT-NLI_large^§	86.10
SRoBERTa-NLI_large^§	86.15
Surrogate-NLI_large	90.69

View Large

This Feature Is Available To Subscribers Only