Machine Translation on Long FLORES en to uk (test)
14.36BLEUGemma 3 12B PT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Gemma 3 12B PT2025.10 | 14.36 | 12.68 | |
| MATTTraining Time=7h 00m2025.10 | 8.7 | 11.81 | |
| Transtokenizers w/ NTPTraining Time=10h 30m2025.10 | 5.13 | 3.24 | |
| Transtokenizers w/ NTPTraining Time=7h 00m2025.10 | 4.63 | 3.14 | |
| FOCUS w/ NTPTraining Time=10h 30m2025.10 | 4.34 | 8.46 | |
| FOCUS w/ NTPTraining Time=7h 00m2025.10 | 4.32 | 8.72 | |
| Transtokenizers w/ NTPTraining Time=3h 30m2025.10 | 3.64 | 2.46 | |
| FOCUS w/ NTPTraining Time=3h 30m2025.10 | 3.51 | 7.56 | |
| FOCUS2025.10 | 1.01 | 2.07 | |
| Transtokenizers2025.10 | 0.04 | 0.06 | |
| WECHSEL2025.10 | 0 | 0 | |
| TokAlign2025.10 | 0 | 0.08 | |
| Token DistillationTraining Time=5h 30m2025.10 | 0 | 6.42 | |
| Token DistillationTraining Time=8h 20m2025.10 | 0 | 6.68 |