Machine Translation Evaluation on WMT 2023 (test)
0.588MAE (EN→DE)Distribution-Calibrated Aggregation
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Distribution-Calibrated Aggregationn (sample count)=12, Base Model=gemini-2.5-flash2025.12 | 0.588 | 0.497 | |
| Distribution-Calibrated Aggregationn (sample count)=4, Base Model=gemini-2.5-flash2025.12 | 0.591 | 0.506 | |
| SCn (sample count)=12, Base Model=gemini-2.5-flash2025.12 | 0.648 | 0.527 | |
| CI-SCn (sample count)=12, Base Model=gemini-2.5-flash2025.12 | 0.652 | 0.524 | |
| Soft-SCn (sample count)=4, Base Model=gemini-2.5-flash2025.12 | 0.664 | 0.557 | |
| CI-SCn (sample count)=4, Base Model=gemini-2.5-flash2025.12 | 0.667 | 0.544 | |
| SCn (sample count)=4, Base Model=gemini-2.5-flash2025.12 | 0.671 | 0.549 | |
| Soft-SCn (sample count)=12, Base Model=gemini-2.5-flash2025.12 | 0.673 | 0.56 | |
| GSCn (sample count)=4, Base Model=gemini-2.5-flash2025.12 | 0.723 | 0.581 | |
| USCn (sample count)=4, Base Model=gemini-2.5-flash2025.12 | 0.728 | 0.527 | |
| GSCn (sample count)=12, Base Model=gemini-2.5-flash2025.12 | 0.755 | 0.546 | |
| USCn (sample count)=12, Base Model=gemini-2.5-flash2025.12 | 0.765 | 0.505 |