SST
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
SST-5 40% symmetric noise BERT-Large (test)
53.1Accuracy (Peak)
9
SST-5 Clean BERT-Large (test)
55.5Accuracy
9
SST-2 (test)
96.4Macro F1
9
SST-2 (test)
1.56AU-MSE
9
SST-2
90.8AUAC
9
SST-5 Root
55.5Accuracy
9
SST-1
48.7Accuracy
9
SST-2
77.6AUROC
8
SST (test)
1.03MSE
8
SST-2
59.32VRAM Usage
8
SST-2 FTC-metadataset mini 10%
0.989AUROC
8
SST-2 FTC-metadataset full
1.132Average Prediction Set Size
8
SST-2 @1024 (test)
79.4Accuracy
8
SST-2
4.4ASR Error
8
SST-2 SuperGLUE (test)
90.78Accuracy
8
SST5 (IID)
0.523Accuracy
8
SST-2 6.7k (test)
89.68Accuracy
8
SST-5
46Avg Acc
8
SST-5 No-Pos
70Positiveness Score
8
SST-5 200-Pos
91Positiveness Score
8
SST-2 20%I (test)
96.08Accuracy
8
SST-5 40%I (test)
55.02Accuracy
8
SST-5 20%I (test)
57.02Accuracy
8
SST-5 40%A (test)
55.39Accuracy
8
SST-5 20%A (test)
57.57Accuracy
8