Machine Translation on WMT De-En 23
29.4BLEUXMOE
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| XMOE#Experts=64, Size=10242024.02 | 29.4 | 53.3 | — | — | — | — | |
| Top-k#Experts=128, Size=5122024.02 | 29.3 | 53.1 | — | — | — | — | |
| Switch#Experts=32, Size=20482024.02 | 29.2 | 53.1 | — | — | — | — | |
| Top-k#Experts=64, Size=10242024.02 | 29.2 | 53.1 | — | — | — | — | |
| XMOE#Experts=128, Size=5122024.02 | 29.2 | 53.2 | — | — | — | — | |
| Dense#Experts=1, Size=20482024.02 | 28.2 | 52.2 | — | — | — | — | |
| OracleDecoding=Oracle, Dist.=-2025.12 | 26.14 | 52.51 | 70.93 | 79.28 | 7.21 | — | |
| MBRDecoding=MBR, Dist.=D62025.12 | 24.28 | 51.37 | 61.8 | 73.42 | 9.38 | 6.97 | |
| AC-PMBRDecoding=AC-PMBR, Dist.=D12, Reduction rate=Low, r=322025.12 | 24.05 | 51.1 | 64.95 | 77.47 | 8.46 | 2.65 | |
| MBRDecoding=MBR, Dist.=D122025.12 | 23.92 | 50.55 | 62.21 | 74.28 | 9.45 | 6.53 | |
| MBRDecoding=MBR, Dist.=-2025.12 | 23.86 | 50.99 | 65.19 | 77.49 | 8.46 | — | |
| AC-PMBRDecoding=AC-PMBR, Dist.=D3, Reduction rate=Low, r=322025.12 | 23.82 | 51.08 | 64.87 | 77.01 | 8.51 | 2.8 | |
| AC-PMBRDecoding=AC-PMBR, Dist.=D6, Reduction rate=Low, r=322025.12 | 23.69 | 50.95 | 64.87 | 77.26 | 8.51 | 2.68 | |
| PMBRDecoding=PMBR, Dist.=-, Reduction rate=Low, r=162025.12 | 23.47 | 50.72 | 64.87 | 77.27 | 8.46 | 2.54 | |
| MBRDecoding=MBR, Dist.=D32025.12 | 23.37 | 51.89 | 60.65 | 70.67 | 9.66 | 8.53 | |
| AC-PMBRDecoding=AC-PMBR, Dist.=D3, Reduction rate=High, r=10242025.12 | 22.96 | 50.5 | 61.08 | 71.98 | 9.71 | 11.89 | |
| AC-PMBRDecoding=AC-PMBR, Dist.=D6, Reduction rate=High, r=10242025.12 | 22.83 | 49.94 | 61.02 | 72.1 | 9.9 | 12.57 | |
| AC-PMBRDecoding=AC-PMBR, Dist.=D12, Reduction rate=High, r=10242025.12 | 22.31 | 49.12 | 61.26 | 72.66 | 9.97 | 16.8 | |
| PMBRDecoding=PMBR, Dist.=-, Reduction rate=High, r=5122025.12 | 21.34 | 48.69 | 60.02 | 70.74 | 10.37 | 33.8 | |
| MAPDecoding=MAP, Dist.=-2025.12 | 16.56 | 37.39 | 56.27 | 65.79 | 11.68 | — |