Model Details bert-large-uncased finetuned on QNLI. Parameter settings batch size is 32, learning rate is 2e-5. Metrics acc: 0.9224
