Question Formation on English Question Formation (Linear generalization)
22.2Full AccuracyTf+Sup+Sup
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Tf+Sup+SupArchitecture=Transformer + Supervision + Supervision2025.11 | 22.2 | 30.3 | |
| Tf+SupArchitecture=Transformer + Supervision2025.11 | 20.2 | 67.5 | |
| TfArchitecture=Transformer2025.11 | 11.2 | 35.5 | |
| Tf+Nd+NdArchitecture=Multi-Stack Attention Transformer2025.11 | 6.7 | 13.5 | |
| Tf+NdArchitecture=Stack Attention Transformer2025.11 | 1.8 | 26.8 |