Machine Translation on Multi-domain Education, Spoken, Thesis, Science, Microblog (test)
36.9BLEU Score (Education)CANEFT
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CANEFTBackbone=Qwen2.5-7B-Instruct, TP (M)=127, AP (M)=02026.02 | 36.9 | 87.1 | 38 | 87.4 | 27.2 | 82.2 | 22.7 | 78.9 | 23.2 | 79.9 | 29.6 | 83.1 | |
| FullBackbone=Qwen2.5-7B-Instruct, TP (M)=7000, AP (M)=02026.02 | 35.1 | 87.2 | 36.8 | 87.3 | 26.6 | 81.3 | 20.6 | 78.3 | 21.5 | 79.2 | 28.1 | 82.6 | |
| CANEFTBackbone=LLaMA3.1-8B-Instruct, TP (M)=146, AP (M)=02026.02 | 34.1 | 86.5 | 34.8 | 86.5 | 25.6 | 81.5 | 18.9 | 77.7 | 22.1 | 79.4 | 27.1 | 82.3 | |
| BaseBackbone=Qwen2.5-7B-Instruct, TP (M)=0, AP (M)=02026.02 | 33.9 | 86.4 | 34.2 | 86.4 | 23.4 | 81.1 | 18.3 | 77.9 | 19.3 | 79.1 | 25.8 | 82.1 | |
| FullBackbone=LLaMA3.1-8B-Instruct, TP (M)=8000, AP (M)=02026.02 | 31.3 | 85.3 | 33.4 | 84.6 | 25.5 | 81.6 | 16.2 | 78.1 | 19.4 | 79.2 | 25.1 | 81.7 | |
| BaseBackbone=LLaMA3.1-8B-Instruct, TP (M)=0, AP (M)=02026.02 | 30.7 | 85.2 | 31.1 | 86 | 21.6 | 80.6 | 15.4 | 77.4 | 17.1 | 79.1 | 23.1 | 81.6 | |
| CANEFTBackbone=LLaMA2-7B-Chat, TP (M)=91, AP (M)=02026.02 | 28.9 | 84.4 | 30.3 | 84.2 | 22.6 | 78.5 | 16.4 | 75.7 | 20.3 | 77.8 | 23.7 | 80.1 | |
| FullBackbone=LLaMA2-7B-Chat, TP (M)=7000, AP (M)=02026.02 | 28.2 | 79.9 | 29.6 | 78.9 | 22.7 | 78.4 | 14.9 | 72.2 | 18.2 | 73.9 | 22.7 | 76.6 | |
| BaseBackbone=LLaMA2-7B-Chat, TP (M)=0, AP (M)=02026.02 | 25.7 | 83.9 | 26.7 | 83.8 | 18.1 | 76.9 | 11.9 | 73.8 | 14.4 | 73.4 | 19.3 | 78.3 |