Question Formation on German question formation
100Full Accuracy (Test)mT5
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| mT5Architecture=mT5, Model Scale=Large pre-trained model2025.11 | 100 | — | — | — | 100 | — | — | |
| mBARTArchitecture=mBART, Model Scale=Large pre-trained model2025.11 | 100 | — | — | — | 82 | — | — | |
| Tf (Ahuja et al. objective)Architecture=Transformer, Training Objective=language modeling2025.11 | 100 | — | — | — | 95 | — | — | |
| Tf+Sup+SupArchitecture=Transformer, Supervision=Double2025.11 | 99.9 | 0 | 0 | 12.645 | 99.6 | 0.4 | 8.688 | |
| Stack Attention Transformer (Tf+Nd)Architecture=Stack Attention Transformer, Training Status=Overtrained2025.11 | 99.9 | 0.4 | 0 | 20.631 | 100 | 0 | 9.81 | |
| Stack Attention Transformer (Tf+Nd+Nd)Architecture=Stack Attention Transformer, Stack Count=22025.11 | 99.7 | 0.1 | 0 | 16.85 | 97.9 | 0 | 8.484 | |
| TfArchitecture=Transformer, Training Status=Overtrained2025.11 | 99.6 | 0 | 0 | 6.535 | 86.8 | 0.3 | 7.33 | |
| Tf+SupArchitecture=Transformer, Supervision=Yes2025.11 | 98.3 | 0 | 0 | 7.458 | 99.9 | 0.1 | 7.251 | |
| Stack LSTM (LSTM+Nd+R)Architecture=Stack LSTM, Randomization=R2025.11 | 64.8 | 0.8 | 0 | 23.347 | 88.7 | 0 | 8.862 | |
| LSTM+Sup+RArchitecture=LSTM, Supervision=Yes, Randomization=R2025.11 | 35.6 | 0.2 | 0 | 22.105 | 91.8 | 7.8 | 6.396 | |
| RNNArchitecture=RNN2025.11 | 32.4 | 0 | 0 | 9.445 | 84.5 | 11 | 2.063 | |
| RNN+SupArchitecture=RNN, Supervision=Yes2025.11 | 27.4 | 0 | 0 | 10.984 | 95.2 | 2.5 | 3.715 | |
| LSTMArchitecture=LSTM, Training Status=Overtrained2025.11 | 23.4 | 1.2 | 0 | 22.493 | 99.2 | 0.3 | 6.823 | |
| RNN+Sup+RArchitecture=RNN, Supervision=Yes, Randomization=R2025.11 | 21 | 0 | 0 | 13.72 | 96.5 | 1.1 | 5.096 | |
| LSTM+SupArchitecture=LSTM, Supervision=Yes2025.11 | 20 | 0.7 | 0 | 18.428 | 98 | 1 | 5.605 | |
| Stack LSTM (LSTM+Nd)Architecture=Stack LSTM2025.11 | 19.9 | 0 | 0 | 22.32 | 96 | 0.3 | 6.758 | |
| Stack RNN (RNN+Nd)Architecture=Stack RNN2025.11 | 18.9 | 0 | 0 | 7.267 | 85.4 | 12.5 | 2.582 | |
| Stack RNN (RNN+Nd+R)Architecture=Stack RNN, Randomization=R2025.11 | 15.3 | 0 | 0 | 10.619 | 89.2 | 9.1 | 2.575 |