Fact Verification on VitaminC (Accuracy and Latency)
54Accuracy (%)CPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CPOBackbone=LLAMA2-13B2024.06 | 54 | 58.6 | |
| CPOBackbone=Mistral-7B2024.06 | 53.7 | 46.6 | |
| CPOBackbone=LLAMA2-7B2024.06 | 52.7 | 40.1 | |
| ToTBackbone=LLAMA2-13B2024.06 | 52.5 | 2,933.6 | |
| ToTBackbone=Mistral-7B2024.06 | 52.2 | 4,698.2 | |
| TS-SFTBackbone=LLAMA2-7B2024.06 | 51 | 40.1 | |
| ToTBackbone=LLAMA2-7B2024.06 | 50.7 | 2,639.3 | |
| TS-SFTBackbone=Mistral-7B2024.06 | 50.3 | 45.9 | |
| TS-SFTBackbone=LLAMA2-13B2024.06 | 49.7 | 51.5 | |
| CoTBackbone=Mistral-7B2024.06 | 47.7 | 44.8 | |
| CoTBackbone=LLAMA2-7B2024.06 | 47.3 | 35.9 | |
| CoTBackbone=LLAMA2-13B2024.06 | 46.3 | 48.3 |