SQuAD
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
SQuAD
29.13EM
10
SQuAD 5-shot
90.38EM Accuracy
10
SQuAD
81.4Recall@1
10
SQuAD trained on CausalQA OOD (test)
27.61F1
10
SQuAD Du
24.43BLEU-4
10
SQuAD 2.0 extracted on Sep 9, 2018 (dev)
86.3EM
10
SQuAD
54.52Accuracy (Clean)
9
SQuAD 2.0
66.4Mean Faithfulness Divergence Reduction
9
SQuAD v1.1
41Hit@1
9
SQuAD
12.23Communication Time (10^3 s)
9
SQuAD
88.9Recall@5
9
SQuAD v1
65.7Exact Match
9
SQuAD
80.08EM Score
9
SQuAD v1.1 (test)
75.1Acc@20
9
SQuAD 1.1 (reversed dev-test)
26.95BLEU-4
9
SQuAD
0.772Reconstruction Rate
8
SQuAD
94.1IE Score
8
SQuAD 2.0
69.08Precision
8
SQuAD
85.25Refusal F1
8
SQuAD Clean (test)
72.84Exact Match (EM)
8
SQuAD
27.79BLEU-4
8
SQuAD Continued Pretraining n=200
46.2Accuracy
8
SQuAD Single Passage n=1
48.9Accuracy
8
SQuAD 2048 tokens
49.4Accuracy
8
SQuAD v1.1
37.8Accuracy
8