Question Answering on MS MARCO (val)
26.7ROUGE-LFocused Hierarchical RNN
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Focused Hierarchical RNN2018.06 | 26.7 | 27.3 | |
| LSTM1 + POINTER SOFTMAX2018.06 | 26.5 | 24.8 | |
| Focused Hierarchical RNNablation_variant=learned boundaries2018.06 | 24 | 23.5 | |
| LSTM2 + POINTER SOFTMAX2018.06 | 23.3 | 24.3 | |
| Focused Hierarchical RNNablation_variant=dot-product between question and context2018.06 | 19.3 | 18.5 | |
| Focused Hierarchical RNNablation_variant=pointer softmax2018.06 | 18.7 | 20.5 | |
| ATTENTION MODEL2018.06 | 12.8 | 9.3 | |
| MEMORY NETWORK2018.06 | 11.9 | — | |
| SEQ-TO-SEQ2018.06 | 8.9 | — |