Idiomatic Translation on Mixed Dataset en-bn
2.25LLM-eval ScoreIdiomCE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| IdiomCEModel=GPT-4o2025.05 | 2.25 | 69 | |
| DirectModel=GPT-4o2025.05 | 1.99 | 76.4 | |
| IdiomCEBackbone=Gemma2-9b-it2025.05 | 1.783 | 72.7 | |
| Indictrans2Inference Method=Direct2025.05 | 1.78 | 81 | |
| IdiomCEModel=Gemma2-9b-it2025.05 | 1.7 | 68 | |
| NLLB-200Inference Method=Direct2025.05 | 1.7 | 78 | |
| DirectBackbone=Gemma2-9b-it2025.05 | 1.49 | 77.5 | |
| IdiomCEBackbone=LLama-3.1-8B2025.05 | 1.46 | 64.04 | |
| DirectModel=Gemma2-9b-it2025.05 | 1.44 | 71 | |
| DirectModel=NLLB-2002025.05 | 1.43 | 76.9 | |
| IdiomCEModel=LLama-3.1-8B2025.05 | 1.4 | 63 | |
| DirectBackbone=LLama-3.1-8B2025.05 | 1.367 | 68.8 | |
| DirectModel=Indictrans22025.05 | 1.275 | 77 | |
| DirectModel=LLama-3.1-8B2025.05 | 1.23 | 67 | |
| IdiomCEBackbone=LLama-3.2-3B2025.05 | 1.23 | 58.67 | |
| IdiomCEModel=LLama-3.2-3B2025.05 | 1.2 | 60 | |
| DirectBackbone=LLama-3.2-3B2025.05 | 1.17 | 61.63 | |
| DirectModel=LLama-3.2-3B2025.05 | 1.05 | 60 |