Machine Translation on Machine Translation Parallel Data (test)
86.69COMETHigh-Quality
Evaluation results
| Method | Links | ||
|---|---|---|---|
| High-QualityModel=LLaMA-3.2-8B-IT, Data Count=50k, Tuning Setting=Transfer-Tuning Scalability2026.01 | 86.69 | 0.1 | |
| MechanisticModel=Gemma-2-9B-IT, Data Count=50k, Tuning Setting=Transfer-Tuning Scalability2026.01 | 86.48 | 0.6 | |
| MechanisticModel=LLaMA-3.2-8B-IT, Data Count=50k, Tuning Setting=Transfer-Tuning Scalability2026.01 | 86.34 | 0.3 | |
| RandomModel=LLaMA-3.2-8B-IT, Data Count=50k, Tuning Setting=Transfer-Tuning Scalability2026.01 | 86.2 | 0.2 | |
| RandomModel=Gemma-2-9B-IT, Data Count=50k, Tuning Setting=Transfer-Tuning Scalability2026.01 | 85.36 | 4.21 | |
| High-LossModel=LLaMA-3.2-8B-IT, Data Count=50k, Tuning Setting=Transfer-Tuning Scalability2026.01 | 84.51 | 0.2 | |
| High-QualityModel=Gemma-2-9B-IT, Data Count=50k, Tuning Setting=Transfer-Tuning Scalability2026.01 | 84.17 | 4.97 | |
| High-LossModel=Gemma-2-9B-IT, Data Count=50k, Tuning Setting=Transfer-Tuning Scalability2026.01 | 83.38 | 8.51 | |
| MechanisticModel=Gemma-2-2B-IT, Data Count=20k, Tuning Setting=Self-Tuning Efficiency2026.01 | 83.37 | 0.9 | |
| High-QualityModel=Gemma-2-2B-IT, Data Count=20k, Tuning Setting=Self-Tuning Efficiency2026.01 | 83.32 | 2.12 | |
| RandomModel=Gemma-2-2B-IT, Data Count=20k, Tuning Setting=Self-Tuning Efficiency2026.01 | 82.49 | 3.62 | |
| High-LossModel=Gemma-2-2B-IT, Data Count=20k, Tuning Setting=Self-Tuning Efficiency2026.01 | 82.14 | 4.32 | |
| OriginalModel=LLaMA-3.2-8B-IT, Tuning Setting=Transfer-Tuning Scalability2026.01 | 80.6 | 10 | |
| OriginalModel=Gemma-2-9B-IT, Tuning Setting=Transfer-Tuning Scalability2026.01 | 79.5 | 12.5 | |
| MechanisticModel=LLaMA-3.1-1B-IT, Data Count=20k, Tuning Setting=Self-Tuning Efficiency2026.01 | 77.92 | 2.39 | |
| High-QualityModel=LLaMA-3.1-1B-IT, Data Count=20k, Tuning Setting=Self-Tuning Efficiency2026.01 | 76.35 | 4.21 | |
| RandomModel=LLaMA-3.1-1B-IT, Data Count=20k, Tuning Setting=Self-Tuning Efficiency2026.01 | 75.88 | 4.62 | |
| OriginalModel=Gemma-2-2B-IT, Tuning Setting=Self-Tuning Efficiency2026.01 | 73.62 | 19.15 | |
| High-LossModel=LLaMA-3.1-1B-IT, Data Count=20k, Tuning Setting=Self-Tuning Efficiency2026.01 | 73.41 | 6.57 | |
| OriginalModel=LLaMA-3.1-1B-IT, Tuning Setting=Self-Tuning Efficiency2026.01 | 57.61 | 32.24 |