Machine Translation on Romansh-to-German Similar Varieties
54.64chrFOur SFT
Evaluation Results
| Method | Links | |
|---|---|---|
| Our SFTBase Model=Qwen3-4B-Base, Training Method=SFT, Context=full2026.06 | 54.64 | |
| Our SFTBase Model=Llama-3.2-3B-Instruct, Training Method=SFT, Context=full2026.06 | 49.7 | |
| Our RLBase Model=Qwen3-4B-Base, Training Method=RL, Context=full2026.06 | 47.85 | |
| Our RLBase Model=Llama-3.2-3B-Instruct, Training Method=RL, Context=full2026.06 | 44.14 | |
| Our SFTBase Model=Qwen3-4B-Base, Training Method=SFT, Context=none2026.06 | 42.35 | |
| Our SFTBase Model=Llama-3.2-3B-Instruct, Training Method=SFT, Context=none2026.06 | 38.26 | |
| Our RLBase Model=Qwen3-4B-Base, Training Method=RL, Context=none2026.06 | 36.94 | |
| Our RLBase Model=Llama-3.2-3B-Instruct, Training Method=RL, Context=none2026.06 | 33.27 | |
| Qwen3-4B-BaseBase Model=Qwen3-4B-Base, Training Method=Base (untuned), Context=full2026.06 | 32.03 | |
| Llama-3.2-3B-InstBase Model=Llama-3.2-3B-Instruct, Training Method=Instruct-tuned, Context=full2026.06 | 28.61 | |
| Qwen3-4B-BaseBase Model=Qwen3-4B-Base, Training Method=Base (untuned), Context=none2026.06 | 24.13 | |
| Llama-3.2-3B-InstBase Model=Llama-3.2-3B-Instruct, Training Method=Instruct-tuned, Context=none2026.06 | 20.37 |