Multilingual Text Generation on Long FLORES and XL-Sum Average
10.02Long FLORES en->x ScoreBaseline
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| BaselineBase Model=Qwen 3 4B2025.10 | 10.02 | 28.29 | 1.93 | 13.41 | |
| BaselineBase Model=Gemma 3 4B PT2025.10 | 7.93 | 20.22 | 4.14 | 10.76 | |
| MATTBase Model=Qwen 3 4B2025.10 | 5.2 | 20.68 | 0.9 | 8.93 | |
| Token DistillationBase Model=Qwen 3 4B2025.10 | 5.11 | 17.74 | 0.04 | 7.63 | |
| FOCUSBase Model=Qwen 3 4B2025.10 | 4.77 | 9.49 | 0.23 | 4.83 | |
| MATTBase Model=Gemma 3 4B PT2025.10 | 3.94 | 13.22 | 1.27 | 6.14 | |
| FOCUSBase Model=Gemma 3 4B PT2025.10 | 1.93 | 8.39 | 0.18 | 3.5 | |
| Token DistillationBase Model=Gemma 3 4B PT2025.10 | 1.31 | 12.87 | 1.09 | 5.09 |