Misclassification Detection on WSJ (Penn Treebank) (test)
0.96AUROCSoftmax Baseline
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Softmax BaselineBackbone=3-layer Bi-LSTM, Neurons per layer=128, Batch size=32, Epochs=10, Optimizer=SGD2016.10 | 0.96 | 1 | 0.51 | 0.71 | 3.68 | |
| Random BaselineDescription=Random guessing baseline2016.10 | 0.5 | 0.96 | 0.037 | — | — |