Machine Translation on FLORES-200 (test)
98.04xCOMET (DE)BridG Post
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| BridG PostMT Model=GPT 3.5, Sentence Bridging Model=Qwen2-72B-Instruct2024.10 | 98.04 | 92.63 | 77.9 | 92.57 | 83.65 | 75.41 | 53.11 | — | |
| BridG PostMT Model=Llama 3.1 70B, Sentence Bridging Model=Qwen2-72B-Instruct2024.10 | 97.93 | — | 84.45 | — | — | — | — | — | |
| BridG Pre & PostMT Model=GPT 3.5, Sentence Bridging Model=Qwen2-72B-Instruct2024.10 | 97.86 | 92.35 | 77.63 | 92.07 | 83.34 | 75.14 | 51.51 | — | |
| BridG Pre & PostMT Model=Llama 3.1 70B, Sentence Bridging Model=Qwen2-72B-Instruct2024.10 | 97.73 | — | 84.31 | — | — | — | — | — | |
| BridG PostMT Model=Mistral Nemo, Sentence Bridging Model=Qwen2-72B-Instruct2024.10 | 97.7 | 90.99 | 73.95 | 89.45 | 45.67 | 78.35 | 57.24 | — | |
| BridG Pre & PostMT Model=Mistral Nemo, Sentence Bridging Model=Qwen2-72B-Instruct2024.10 | 97.7 | 90.99 | 73.82 | 89.34 | 45.18 | 77.83 | 54.55 | — | |
| TowerInstructModel Size=13B2024.10 | 97.69 | 89.89 | — | 91.29 | — | — | — | — | |
| MAPSMT Model=GPT 3.52024.10 | 97.68 | 92.06 | — | — | — | — | — | — | |
| Zero-shotMT Model=GPT 3.52024.10 | 97.63 | 91.3 | 71.89 | 89.48 | 81.23 | 68.73 | 44.53 | — | |
| QA DecodeMT Model=GPT 3.52024.10 | 97.45 | 91.63 | 75.45 | 90.19 | 78.26 | 73.9 | 45.82 | — | |
| QA DecodeMT Model=Llama 3.1 70B2024.10 | 97.37 | — | — | — | — | — | — | — | |
| Zero-shotMT Model=Llama 3.1 70B2024.10 | 97.33 | — | 79.41 | — | — | — | — | — | |
| BridG PostMT Model=Llama 3.1 8B, Sentence Bridging Model=Qwen2-72B-Instruct2024.10 | 97.25 | — | 78.3 | — | — | — | — | — | |
| BridG Pre & PostMT Model=Llama 3.1 8B, Sentence Bridging Model=Qwen2-72B-Instruct2024.10 | 96.99 | — | 77.37 | — | — | — | — | — | |
| Zero-shotMT Model=Mistral Nemo2024.10 | 96.7 | 88.01 | 66.78 | 81.43 | 38.97 | 71.89 | 43.99 | — | |
| MAPSMT Model=Llama 3.1 70B2024.10 | 96.53 | — | — | — | — | — | — | — | |
| QA DecodeMT Model=Mistral Nemo2024.10 | 96.33 | 87.18 | 64.42 | 80.34 | 28.86 | 68.59 | 35.37 | — | |
| NLLBModel Version=NLLB-200-3.3B2024.10 | 96.21 | 67.88 | 81 | 82.2 | 77.17 | 82.7 | 71.83 | — | |
| Zero-shotMT Model=Llama 3.1 8B2024.10 | 94.99 | — | 69.93 | — | — | — | — | — | |
| QA DecodeMT Model=Llama 3.1 8B2024.10 | 88.08 | — | 77.67 | — | — | — | — | — | |
| MAPSMT Model=Mistral Nemo2024.10 | 85.75 | 88.71 | — | — | — | — | — | — | |
| MAPSMT Model=Llama 3.1 8B2024.10 | 79.1 | — | — | — | — | — | — | — | |
| Gemma-3-1b-ittuning=Instruction-tuned, prompt_language=English2025.12 | — | — | — | — | — | — | — | 49.4 | |
| MiniLingua-1b-Instructtuning=Instruction-tuned, prompt_language=English2025.12 | — | — | — | — | — | — | — | 68.1 | |
| SmolLM2-1.7b-Instructtuning=Instruction-tuned, prompt_language=English2025.12 | — | — | — | — | — | — | — | 49.6 |