Machine Translation on WMT ZH-EN 24
81.06COMETkiwiClaude-3.5-Sonnet
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Claude-3.5-Sonnet2025.05 | 81.06 | — | 90.54 | |
| SSR-X-Zero-7BReasoning/Thinking Mode=true2025.05 | 80.56 | — | 89.42 | |
| Qwen2.5-72B-Instruct2025.05 | 80.46 | — | 90.06 | |
| Qwen3-32BReasoning/Thinking Mode=true2025.05 | 79.95 | — | 89.53 | |
| GPT-4o2025.05 | 79.9 | — | 89.06 | |
| Qwen3-8BReasoning/Thinking Mode=true2025.05 | 79.58 | — | 89.02 | |
| Gemma2-27B-it2025.05 | 79.42 | — | 89.14 | |
| Gemma2-9B-it2025.05 | 79.25 | — | 88.41 | |
| Qwen3-32B2025.05 | 79.2 | — | 88.47 | |
| SSR-Zero-7B2025.05 | 79.04 | — | 89.19 | |
| Gemini-1.5-Pro2025.05 | 79.02 | — | 88.9 | |
| Qwen2.5-32B-Instruct2025.05 | 78.77 | — | 88.69 | |
| GemmaX2-28-9B-v0.12025.05 | 78.71 | — | 88.6 | |
| Qwen3-8B2025.05 | 77.99 | — | 86.94 | |
| SSR-Zero-3B2025.05 | 77.81 | — | 86.18 | |
| TowerInstruct-13B-v0.12025.05 | 77.57 | — | 87.12 | |
| TowerInstruct-7B-v0.22025.05 | 76.96 | — | 85.98 | |
| Qwen2.5-7B-Instruct2025.05 | 76.71 | — | 87.12 | |
| Qwen2.5-3B-Instruct2025.05 | 75.82 | — | 85.03 | |
| QwQ-32BReasoning/Thinking Mode=true2025.05 | 75.08 | — | 84.34 | |
| Qwen2.5-7B2025.05 | 69.04 | — | 77.33 | |
| MT-R1-ZeroBackbone=Qwen3-0.6B, RL training=MT-R1-Zero, Rollouts per example=722026.02 | 66.86 | 15.97 | 77.74 | |
| PEGRLBackbone=Qwen3-0.6B, RL training=PEGRL, Rollouts per example=722026.02 | 66.69 | 16.26 | 78.28 | |
| Qwen3-0.6BTraining=Pre-trained only2026.02 | 63.87 | 15 | 75.62 | |
| Qwen2.5-3B2025.05 | 41.66 | — | 55.16 |