Machine Translation on WMT ET-EN (test)
16.8BLEUMix-MoE
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Mix-MoEBase Model=Llama3.2-1B, Training Strategy=Mix-MoE2026.05 | 16.8 | 38.6 | 67.1 | |
| MOE-LPRBase Model=Llama3.2-1B, Training Strategy=MOE-LPR2026.05 | 16.3 | 37.2 | 66.6 | |
| Full FinetuneBase Model=Llama3.2-1B, Training Strategy=Full Finetune2026.05 | 15.4 | 36.3 | 66.9 | |
| Dense PT+SFTBase Model=Llama3.2-1B, Training Strategy=Dense PT+SFT2026.05 | 14.9 | 37 | 67 | |
| MoEBase Model=Llama3.2-1B, Training Strategy=MoE2026.05 | 13.3 | 32.7 | 56.9 | |
| LLAMA-ProBase Model=Llama3.2-1B, Training Strategy=LLAMA-Pro2026.05 | 11.5 | 31 | 60.8 | |
| LoRABase Model=Llama3.2-1B, Training Strategy=LoRA2026.05 | 10 | 27.9 | 59.5 | |
| LLAMA-MOE v2Base Model=Llama3.2-1B, Training Strategy=LLAMA-MOE v22026.05 | 8.6 | 19.5 | 60.6 | |
| Qwen2.5-1.5BBase Model=Qwen2.5-1.5B2026.05 | 7 | 19.4 | 59.9 | |
| Qwen2.5-0.5BBase Model=Qwen2.5-0.5B2026.05 | 3.8 | 10.6 | 52.1 | |
| Llama3.2-1BBase Model=Llama3.2-1B2026.05 | 2.4 | 4.1 | 42.2 |