GLUE
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
GLUE BERT-base experts (test)
60.7RTE Score
6
GLUE
90.4Accuracy (SST-2)
6
GLUE
50.21NLI Accuracy
6
GLUE benchmark v1 (test)
72.01COLA Score
6
GLUE
0.698CoLA Score
5
GLUE (test)
91.55MNLI Score
5
GLUE Dirichlet label skew α=0.5 (Dir(0.5))
85.71MNLI-m Accuracy
5
GLUE Dirichlet label skew α=0.1 (Dir(0.1))
72.89MNLI-m Accuracy
5
GLUE
1Coverage Loss
5
GLUE
62.33CoLA Score
5
GLUE (test)
86.8MRPC Score
5
GLUE
57.11SST-2 Accuracy
5
GLUE
92.9SST-2 Accuracy
5
GLUE (val)
94.61SST-2 Accuracy
5
GLUE (test)
92.6STS-B Score
5
GLUE (QNLI, QQP, MNLI) (test)
87.9QNLI Score
5
GLUE
63.3CoLA Score
4
GLUE
12.4CoLA Score
4
GLUE (imbalanced partition)
94.725SST-2 Accuracy
4
GLUE
86.33MNLI Accuracy
4
GLUE QNLI, MNLI, SST-2, QQP
62QNLI Accuracy
4
GLUE (val)
87.5SST-2 Accuracy
4
GLUE
84.98MRPC F1
4
GLUE downstream fine-tuning
31.5CoLA Score
4
GLUE excluding STS-B
88.4Average Score
4