PAWS
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
PAWS
0.37G(W) Error Approximation
40
PAWS
91.24AUROC
36
PAWS
97.6Accuracy
35
PAWS
94.6Accuracy
24
PAWS original (test)
91.79Accuracy
23
PAWS -> PAWS (test)
94.1Accuracy
22
PAWS -> QQP (test)
77.7AUROC
20
PAWS-QQP
96Accuracy
16
PAWS Wiki
86.8Accuracy
12
PAWS
1.98Generalization Error G(W)
10
PAWS Out-of-distribution from PIT
55.1Macro F1
10
PAWS -> WMT (test)
0.849AUROC
10
PAWS Wiki
0.96Kendall Correlation
10
PAWS
0.8MAE
9
PAWS
1.92MAE
9
PAWS QQP
0.87Kendall Correlation
7
PAWS (test)
94.77Accuracy
4
PAWS
56.1Accuracy (Normalized)
3