Machine Translation on WMT23 (test)
83.2COMET ScoreX-ALMA-13B (Only SFT, MoE)
Evaluation Results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| X-ALMA-13B (Only SFT, MoE)# Tokens (Pre-training)=110B, # Params (Base/Adapter)=13B/16B, # Params (Total)=29B2025.02 | 83.2 | — | — | — | — | — | — | — | — | — | — | — | — | |
| X-ALMA-13B-DAT (MoE)# Tokens (Pre-training)=20B, # Params (Base/Adapter)=13B/4B, # Params (Total)=17B2025.02 | 82.8 | — | — | — | — | — | — | — | — | — | — | — | — | |
| Aya-23-35B# Tokens (Pre-training)=null, # Params (Base/Adapter)=35B/-, # Params (Total)=35B2025.02 | 82.3 | — | — | — | — | — | — | — | — | — | — | — | — | |
| X-ALMA-13B-DATM (MoE)# Tokens (Pre-training)=20B, # Params (Base/Adapter)=13B/1B, # Params (Total)=14B2025.02 | 82.1 | — | — | — | — | — | — | — | — | — | — | — | — | |
| Aya-23-8B# Tokens (Pre-training)=null, # Params (Base/Adapter)=8B/-, # Params (Total)=8B2025.02 | 81.5 | — | — | — | — | — | — | — | — | — | — | — | — | |
| LLaMAX3-Alpaca-8B# Tokens (Pre-training)=66B, # Params (Base/Adapter)=8B/-, # Params (Total)=8B2025.02 | 81 | — | — | — | — | — | — | — | — | — | — | — | — | |
| Aya-101# Tokens (Pre-training)=null, # Params (Base/Adapter)=13B/-, # Params (Total)=13B2025.02 | 79.7 | — | — | — | — | — | — | — | — | — | — | — | — | |
| NLLB-3.3B# Tokens (Pre-training)=null, # Params (Base/Adapter)=3B/-, # Params (Total)=3B2025.02 | 71.2 | — | — | — | — | — | — | — | — | — | — | — | — | |
| Aya 23 35BParameters=35B2025.02 | — | 23.5 | 79.7 | 32.7 | 82.3 | 31.7 | 82.2 | 21.3 | 81.6 | 39.1 | 85.7 | 29.7 | 82.3 | |
| Aya 23 8BParameters=8B2025.02 | — | 22.6 | 78.8 | 32.3 | 82.1 | 30.9 | 81.6 | 19.8 | 80.2 | 39.2 | 85 | 29 | 81.5 | |
| Aya-101Parameters=13B2025.02 | — | 13.8 | 73.7 | 34.9 | 81.6 | 28.4 | 81.4 | 13.9 | 77.3 | 34.9 | 84.5 | 25.2 | 79.7 | |
| LLaMAX3-Alpaca-8BBackbone=LLaMA-3, Parameters=8B2025.02 | — | 22.3 | 79.3 | 25.6 | 79.4 | 29.4 | 81.3 | 17.6 | 80.1 | 37.8 | 84.9 | 26.5 | 81 | |
| NLLB-3.3BParameters=3.3B2025.02 | — | 11.4 | 67.8 | 20.1 | 66.6 | 24.4 | 76.7 | 6.8 | 65.8 | 33.1 | 79 | 19.2 | 71.2 | |
| X-ALMA-13B (Only SFT, MoE)Architecture=MoE, Parameters=13B, Training Mode=Only SFT2025.02 | — | 23.8 | 80.3 | 42.5 | 85.3 | 32.8 | 82.4 | 20.4 | 81.6 | 42.5 | 86.4 | 32.4 | 83.2 | |
| X-ALMA-13B-DAT (MoE)Architecture=MoE, Parameters=13B, Training Mode=DAT2025.02 | — | 21.1 | 79.7 | 38.4 | 84.6 | 32.3 | 82.4 | 19.8 | 81.2 | 39.5 | 85.9 | 30.2 | 82.8 | |
| X-ALMA-13B-DATM (MoE)Architecture=MoE, Parameters=13B, Training Mode=DATM2025.02 | — | 22.2 | 79.2 | 37.5 | 84.2 | 29 | 82 | 19 | 80.4 | 34 | 84.7 | 28.3 | 82.1 |