Multi-modal Machine Translation on Multi30k WMT17 (test)
42BLEUGoogle Translate
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Google Translate2022.12 | 42 | — | |
| Plain TransformerTraining Data=+PT2022.12 | 40.46 | — | |
| FUSION+PROMPTTraining Data=+PT2022.12 | 40.07 | — | |
| FUSION+PROMPTTraining Data=Triplet Only2022.12 | 34.62 | — | |
| UPOC2Training Data=Triplet Only2022.12 | 34.1 | — | |
| FUSION-BASEDTraining Data=Triplet Only2022.12 | 34.03 | — | |
| PROMPT-BASEDTraining Data=Triplet Only2022.12 | 33.87 | — | |
| Selective AttentionTraining Data=Triplet Only2022.12 | 33.8 | — | |
| Phrase RetrievalTraining Data=Triplet Only2022.12 | 33.45 | — | |
| Plain TransformerTraining Data=Triplet Only2022.12 | 31.87 | — | |
| UVR-NMTTraining Data=Triplet Only2022.12 | 31.85 | — | |
| Gumbel-Attention MMT2021.03 | 31.4 | 51.2 | |
| Trg-mul2021.03 | 30.7 | 52.2 | |
| Latent Variable MMT2021.03 | 30.1 | 49.9 | |
| Fusion-conv2021.03 | 29.8 | 51.2 | |
| Text-only Transformer2021.03 | 29.1 | 48.6 |