Machine Translation on FLORES Zh-En
26.25BLEUTIM-7B
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| TIM-7B2024.05 | 26.25 | 85.81 | — | — | — | |
| G-DIGDiversity=false2024.05 | 25.81 | 86.45 | 75.19 | — | — | |
| G-DIGDiversity=true2024.05 | 25.04 | 86.18 | 74.75 | — | — | |
| Reward2024.05 | 24.48 | 86.16 | 74.73 | — | — | |
| Random2024.05 | 24.25 | 85.63 | 73.84 | — | — | |
| PEGRLBackbone=Qwen3-0.6B, RL training=PEGRL, Rollouts per example=722026.02 | 20.68 | — | — | 75.49 | 90.48 | |
| MT-R1-ZeroBackbone=Qwen3-0.6B, RL training=MT-R1-Zero, Rollouts per example=722026.02 | 19.66 | — | — | 74.91 | 89.69 | |
| Qwen3-0.6BTraining=Pre-trained only2026.02 | 19.32 | — | — | 72.85 | 88.34 |