Skip to Main Content
Table 6: 

RoBERTa-base and SIFT’s performance on the entire MNLI development sets and their absolute and relative differences, with different numbers of finetuning instances randomly subsampled from the training data.

ID.OOD.
Fraction|Train|RoBERTaSIFTAbsΔRelΔRoBERTaSIFTAbsΔRelΔ
100% 392k 87.7 87.9 0.2 0.2% 87.3 87.7 0.4 0.4% 
0.5% 1,963 76.1 77.6 1.5 1.9% 77.1 78.2 1.1 1.4% 
0.2% 785 68.6 71.0 2.5 3.5% 70.0 71.8 1.8 2.5% 
0.1% 392 58.7 61.2 2.6 4.2% 60.5 63.7 3.3 5.1% 
ID.OOD.
Fraction|Train|RoBERTaSIFTAbsΔRelΔRoBERTaSIFTAbsΔRelΔ
100% 392k 87.7 87.9 0.2 0.2% 87.3 87.7 0.4 0.4% 
0.5% 1,963 76.1 77.6 1.5 1.9% 77.1 78.2 1.1 1.4% 
0.2% 785 68.6 71.0 2.5 3.5% 70.0 71.8 1.8 2.5% 
0.1% 392 58.7 61.2 2.6 4.2% 60.5 63.7 3.3 5.1% 
Close Modal

or Create an Account

Close Modal
Close Modal