List your company
ResearchFormerly WizWand

Machine Translation on English-to-Dinka

22.91chrFOur RL
2.50527.802613.118.3974Jun 4, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Our RLBase Model=Qwen3-4B-Base, Training Method=RL, Context=full2026.0622.91
Our RLBase Model=Llama-3.2-3B-Instruct, Training Method=RL, Context=full2026.0619.49
Qwen3-4B-BaseBase Model=Qwen3-4B-Base, Training Method=Base (untuned), Context=full2026.0616.06
Qwen3-4B-BaseBase Model=Qwen3-4B-Base, Training Method=Base (untuned), Context=none2026.0614.52
Our SFTBase Model=Qwen3-4B-Base, Training Method=SFT, Context=none2026.0610.75
Llama-3.2-3B-InstBase Model=Llama-3.2-3B-Instruct, Training Method=Instruct-tuned, Context=full2026.0610.44
Our RLBase Model=Llama-3.2-3B-Instruct, Training Method=RL, Context=none2026.069.68
Our RLBase Model=Qwen3-4B-Base, Training Method=RL, Context=none2026.066.43
Llama-3.2-3B-InstBase Model=Llama-3.2-3B-Instruct, Training Method=Instruct-tuned, Context=none2026.066.4
Our SFTBase Model=Qwen3-4B-Base, Training Method=SFT, Context=full2026.065.06
Our SFTBase Model=Llama-3.2-3B-Instruct, Training Method=SFT, Context=none2026.064.1
Our SFTBase Model=Llama-3.2-3B-Instruct, Training Method=SFT, Context=full2026.063.29