Machine Translation on Reference-based Mongolian-to-Korean Translation Dataset
14.64BLEU-3CoPiT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CoPiTBackbone=Ministral-3 (14B), Evaluation Protocol=Fine-tuned2026.07 | 14.64 | 9.82 | 16.01 | 15.3 | 0.714 | |
| CoPiTBackbone=GPT-4.1, Evaluation Protocol=Zero-shot2026.07 | 13.37 | 9.04 | 14.59 | 14.05 | 0.667 | |
| CoPiTBackbone=Qwen-3 (4B), Evaluation Protocol=Fine-tuned2026.07 | 10.78 | 6.65 | 11.15 | 10.91 | 0.639 | |
| CoPiTBackbone=Qwen-3 (30B), Evaluation Protocol=Fine-tuned2026.07 | 9.31 | 5.81 | 10.91 | 10.65 | 0.595 | |
| CoPiTBackbone=Ministral-3 (3B), Evaluation Protocol=Fine-tuned2026.07 | 5.59 | 3.17 | 9.46 | 8.88 | 0.585 | |
| CoPiTBackbone=Ministral-3 (14B), Evaluation Protocol=Zero-shot2026.07 | 3.37 | 1.8 | 6.93 | 6.72 | 0.528 | |
| CoPiTBackbone=Qwen-3 (30B), Evaluation Protocol=Zero-shot2026.07 | 2.95 | 1.43 | 6.08 | 5.87 | 0.498 | |
| DirectBackbone=GPT-4.1, Evaluation Protocol=Direct2026.07 | 2.67 | 1.39 | 5.3 | 5.04 | 0.454 | |
| DirectBackbone=Ministral-3 (14B), Evaluation Protocol=Direct2026.07 | 1.44 | 0.65 | 4.53 | 4.3 | 0.435 | |
| DirectBackbone=Qwen-3 (30B), Evaluation Protocol=Direct2026.07 | 1.28 | 0.59 | 3.94 | 3.85 | 0.421 | |
| CoPiTBackbone=Qwen-3 (4B), Evaluation Protocol=Zero-shot2026.07 | 0.92 | 0.41 | 3.5 | 3.36 | 0.372 | |
| CoPiTBackbone=Ministral-3 (3B), Evaluation Protocol=Zero-shot2026.07 | 0.81 | 0.3 | 3.61 | 3.42 | 0.393 | |
| DirectBackbone=Qwen-3 (4B), Evaluation Protocol=Direct2026.07 | 0.74 | 0.35 | 2.88 | 2.75 | 0.351 | |
| DirectBackbone=Ministral-3 (3B), Evaluation Protocol=Direct2026.07 | 0.39 | 0.25 | 0.65 | 0.68 | 0.236 |