Text Classification on QNLI (test)
92.8Accuracy (Clean)Fine-Tuned
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Fine-TunedPLM=RoBERTa-base2024.06 | 92.8 | 26.6 | 22.5 | 20.3 | 204 | 219 | 188 | |
| FreeLB++PLM=RoBERTa-base2024.06 | 92.8 | 33.9 | 27.7 | 20 | 227 | 244 | 200 | |
| DiffuseDef-FreeLB++PLM=RoBERTa-base2024.06 | 92.7 | 64.6 | 64.3 | 61.5 | 473 | 579 | 524 | |
| InfoBERTPLM=RoBERTa-base2024.06 | 92.5 | 29.1 | 25.8 | 20.3 | 205 | 223 | 189 | |
| InfoBERTBackbone=BERT2024.06 | 91.2 | 27.8 | 22.8 | 18.7 | 217 | 232 | 201 | |
| Fine-TunedBackbone=BERT2024.06 | 90.8 | 21.5 | 15.5 | 13.3 | 195 | 206 | 177 | |
| ATINTERBackbone=BERT2024.06 | 90.4 | 43 | 24.8 | 34 | 185 | 115 | 89 | |
| FreeLB++Backbone=BERT2024.06 | 90.3 | 45.6 | 40.2 | 30.5 | 253 | 279 | 226 | |
| DiffuseDef-FreeLB++Backbone=BERT2024.06 | 90.3 | 66.7 | 65.3 | 64.4 | 485 | 587 | 543 | |
| RSMIPLM=RoBERTa-base2024.06 | 89.3 | 38.8 | 33.2 | 30 | 340 | 375 | 343 | |
| EarlyRobustBackbone=BERT2024.06 | 89.2 | 24.3 | 21.2 | 19.1 | 265 | 292 | 255 | |
| DiffuseDef-RSMIPLM=RoBERTa-base2024.06 | 88.8 | 57.7 | 55.7 | 53.9 | 469 | 578 | 518 | |
| RSMIBackbone=BERT2024.06 | 87.4 | 35.2 | 30.9 | 28.2 | 314 | 353 | 314 | |
| DiffuseDef-RSMIBackbone=BERT2024.06 | 86.4 | 55.5 | 54.8 | 57.2 | 459 | 569 | 528 |