Average Binary Classification on QQP, QNLI, and MRPC
81.51Average AUCMixSP
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MixSPBackbone=SimCSE-BERT-Base, Modification=MixSP2024.06 | 81.51 | 82.08 | 80.14 | 82.32 | |
| MixSPBackbone=DiffAug-BERT-Base, Modification=MixSP2024.06 | 81.45 | 81.85 | 80.11 | 82.4 | |
| MixSPBackbone=SBERT-BERT-Base, Modification=MixSP2024.06 | 81.24 | 82.29 | 78.21 | 83.21 | |
| MoEBackbone=SimCSE-BERT-Base, Modification=MoE2024.06 | 81.19 | 81.66 | 79.36 | 82.54 | |
| MoEBackbone=DiffAug-BERT-Base, Modification=MoE2024.06 | 80.87 | 81.33 | 78.64 | 82.63 | |
| MoEBackbone=SBERT-BERT-Base, Modification=MoE2024.06 | 80.53 | 81.6 | 76.91 | 83.08 | |
| TENC-mutual (cross)Backbone=BERT-base, Encoder Architecture=cross-encoder, Training Strategy=TENC-mutual2021.09 | 79.72 | — | — | — | |
| TENC-mutual (bi)Backbone=BERT-base, Encoder Architecture=bi-encoder, Training Strategy=TENC-mutual2021.09 | 79.18 | — | — | — | |
| SimCSE + FTBackbone=SimCSE-BERT-Base, Modification=Fine-Tuning2024.06 | 78.81 | 82.55 | 72.4 | 81.48 | |
| DiffAugBackbone=DiffAug-BERT-Base2024.06 | 78.05 | 81.9 | 74.17 | 78.08 | |
| TENC (cross)Backbone=BERT-base, Encoder Architecture=cross-encoder, Training Strategy=TENC2021.09 | 77.97 | — | — | — | |
| SimCSEBackbone=SimCSE-BERT-Base2024.06 | 77.91 | 81.37 | 74.53 | 77.83 | |
| TENC (bi)Backbone=BERT-base, Encoder Architecture=bi-encoder, Training Strategy=TENC2021.09 | 77.7 | — | — | — | |
| DiffAug + FTBackbone=DiffAug-BERT-Base, Modification=Fine-Tuning2024.06 | 77.27 | 81.39 | 69.11 | 81.3 | |
| SBERTBackbone=SBERT-BERT-Base2024.06 | 76.36 | 80.44 | 68.51 | 80.13 | |
| SimCSE-BERT-baseBackbone=BERT-base2021.09 | 75.59 | — | — | — |