Question Answering on SQuAD (v1.1 & v2.0 EM/F1)
80.22F1 Score (SQuAD v1.1)BoT
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| BoTArchitecture=BERT-B, Evaluation Protocol=Directly fine-tuned2026.03 | 80.22 | 70.34 | 40.99 | 34.5 | 56.51 | |
| MangoArchitecture=BERT-B, Evaluation Protocol=Directly fine-tuned2026.03 | 79.24 | 69.51 | 39.53 | 34.27 | 55.64 | |
| LiGOArchitecture=BERT-B, Evaluation Protocol=Directly fine-tuned2026.03 | 79.23 | 69.65 | 39.45 | 34.42 | 55.69 | |
| BoTArchitecture=BERT-S, Evaluation Protocol=Directly fine-tuned2026.03 | 69.63 | 57.87 | 35.77 | 29.71 | 48.25 | |
| WSArchitecture=BERT-S, Evaluation Protocol=Directly fine-tuned2026.03 | 32.99 | 23.38 | 17.11 | 11.65 | 21.28 | |
| bert2BERTArchitecture=BERT-B, Evaluation Protocol=Directly fine-tuned2026.03 | 31.58 | 22.2 | 16.25 | 11.2 | 20.31 | |
| ScratchArchitecture=BERT-B, Evaluation Protocol=Directly fine-tuned2026.03 | 18.95 | 9.62 | 9.23 | 4.55 | 10.59 | |
| ScratchArchitecture=BERT-S, Evaluation Protocol=Directly fine-tuned2026.03 | 18.28 | 9.05 | 8.13 | 3.72 | 9.8 |