Natural Language Understanding on GLUE (SST-2, CoLA, MRPC, QNLI Subset)
92.55SST-2 AccuracyPyTorch
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| PyTorchModel=BERT-large, Arithmetic Semantics=Floating-point2026.06 | 92.55 | 85.52 | 87.74 | 92.49 | |
| FuseFSSModel=BERT-large, Arithmetic Semantics=Fixed-point2026.06 | 92.5 | 85.57 | 87.5 | 92.66 | |
| PyTorchModel=BERT-base, Arithmetic Semantics=Floating-point2026.06 | 89.33 | 83.43 | 88.73 | 91.55 | |
| FuseFSSModel=BERT-base, Arithmetic Semantics=Fixed-point2026.06 | 89.33 | 83.45 | 88.48 | 91.67 | |
| PyTorchModel=BERT-tiny, Arithmetic Semantics=Floating-point2026.06 | 80.39 | — | 76.37 | 85.69 | |
| FuseFSSModel=BERT-tiny, Arithmetic Semantics=Fixed-point2026.06 | 80.39 | — | 76.96 | 86.23 |