Spoken Language Understanding on SILICONE 1.0 (test)
74.3Avg ScoreHT
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| HTPre-training Objective=MLM, Model Scale=SMALL, Supervision Level=Utterance and Dialog (u_d), Decoder=MLP2020.09 | 74.3 | 79.2 | 92.4 | 81.5 | 90.6 | 69.4 | 92.7 | 64.1 | 60.1 | 45 | 68.2 | |
| HTPre-training Objective=MLM, Model Scale=TINY, Supervision Level=Utterance and Dialog (u_d), Decoder=MLP2020.09 | 73.3 | 79.3 | 92 | 80.1 | 90 | 68.3 | 92.5 | 62.6 | 59.9 | 42 | 66.6 | |
| BERTDecoder=MLP2020.09 | 72.8 | 79.2 | 90.7 | 82.6 | 88.2 | 66.9 | 91.9 | 59.3 | 61.4 | 45 | 62.7 | |
| HTPre-training Objective=GAP, Model Scale=TINY, Supervision Level=Utterance and Dialog (u_d), Decoder=MLP2020.09 | 71.6 | 78.6 | 91.8 | 78.1 | 89.3 | 64.1 | 91.6 | 60.5 | 55.7 | 42.2 | 63.9 | |
| BERTModel Scale=4 layers, Decoder=MLP2020.09 | 70.4 | 77.8 | 90.7 | 79 | 88.4 | 66.8 | 90.3 | 55.3 | 53.4 | 43 | 58.8 | |
| HRDecoder=MLP2020.09 | 69.8 | 77.5 | 90.9 | 80.1 | 82.8 | 64.3 | 91.5 | 59.3 | 59.9 | 40.3 | 51.1 |