Natural Language Understanding on GLUE (MRPC, SST-2, CoLA Subset)
84.98MRPC F1BERT-Base
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| BERT-BaseMode=Fine-tuning, Backbone=Frozen, Classifier Head=768-dim hidden layer with Tanh activation2026.02 | 84.98 | 92.43 | 56.78 | 78.06 | |
| TernaryLMMode=Fine-tuning, Backbone=Frozen quantized, Precision=1-bit, Classifier Head=768-dim hidden layer with Tanh activation2026.02 | 82.47 | 88.92 | 47.23 | 72.87 | |
| DistilBERTMode=Fine-tuning, Backbone=Frozen, Classifier Head=768-dim hidden layer with Tanh activation2026.02 | 82.31 | 91.12 | 51.32 | 74.92 | |
| TinyBERTMode=Fine-tuning, Backbone=Frozen, Classifier Head=768-dim hidden layer with Tanh activation2026.02 | 80.45 | 89.67 | 48.91 | 73.01 |