Machine Translation on FLORES-200 (hye->en, test)
88.8COMET-22SCALE-refine
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SCALE-refineLLM=GPT-4, STM=NLLB-3.3B2023.09 | 88.8 | 88 | 77.8 | 42.3 | |
| NLLBModel Version=3.3B, Evaluation Protocol=Supervised2023.09 | 88.3 | 87.5 | 77 | 43 | |
| GPT-4Evaluation Protocol=few-shot2023.09 | 86.2 | 86 | 73.1 | 35.6 | |
| MicrosoftType=Commercial System2023.09 | 85.2 | 85 | 71.5 | 34.6 | |
| GPT-3.5Evaluation Protocol=few-shot2023.09 | 77 | 77.2 | 60.5 | 19.4 | |
| M2M100Type=Specialized Model2023.09 | 75.9 | 76.5 | 58.9 | 23.7 | |
| XGLMEvaluation Protocol=few-shot2023.09 | 43.9 | 50.2 | 20.5 | 0.2 |