Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

CoLA

Benchmarks

Task NameDataset NameSOTA ResultTrend
Linguistic AcceptabilityCoLA
Accuracy (CoLA)85.91
108
Linguistic AcceptabilityCoLA (test)
Avg Accuracy86.34
56
Linguistic AcceptabilityCoLA
Generalization Error Approximation (G(W))0.25
40
Text Reconstruction AttackCoLA
Total Runtime (hours)0.1
36
Membership Inference AttackCoLA
AUC100
32
Sequence reconstructionCoLA
ROUGE-199.6
32
Training Data ReconstructionCoLA
ROUGE-1100
32
Misclassification DetectionCola
ROC-AUC81.6
31
Linguistic AcceptabilityCOLA
Accuracy83.43
31
Text reconstruction from gradientsCoLA
ROUGE-1100
24
Acceptability ClassificationCoLA out-of-domain (dev test)
Accuracy84.9
24
Text Reconstruction from GradientsCoLA
Average per-batch runtime (seconds)3
20
Text ClassificationCoLA (test)
MCC61.21
18
Natural Language UnderstandingCoLA
Exact Match77.7
14
Grammatical AcceptabilityCoLA (held-out)
F1 Score46.8
14
Text-to-Image retrievalCola (test)
Multi-Obj Acc83.88
14
Dependency tree recoveryCoLA max length 14 (val)
ARR75.7
14
Linguistic Acceptability ClassificationCoLA
Utility Accuracy85
12
Text ReconstructionCoLA
ROUGE-L96.3
12
Acceptability ClassificationCoLA English (IDD)
Accuracy88.2
12
Acceptability ClassificationCoLA in-domain (dev)
Accuracy88.6
12
Natural Language InferenceCoLA
Accuracy86.26
10
Compositional ReasoningCola
Txt2Img Score33.33
10
Linguistic AcceptabilityCoLA GLUE (val)
Accuracy84.76
9
Sentence ClassificationCoLA full (test)
Accuracy86.4
9
Showing 25 of 37 rows