Factuality Evaluation on FactScore (unlabeled)
76.4US (%)PaCE
Evaluation results
| Method | Links | ||
|---|---|---|---|
| PaCETarget Model=LLaMA2-13B-Chat2024.06 | 76.4 | 55.1 | |
| PromptingTarget Model=LLaMA2-13B-Chat2024.06 | 47.5 | 20 | |
| OrthoProjTarget Model=LLaMA2-13B-Chat2024.06 | 43.2 | 51.7 | |
| VanillaTarget Model=LLaMA2-13B-Chat2024.06 | 41.8 | 38.5 | |
| PaCETarget Model=LLaMA2-7B-Chat2024.06 | 30.8 | 73.3 | |
| OrthoProjTarget Model=LLaMA2-7B-Chat2024.06 | 26.2 | 45.9 | |
| PromptingTarget Model=LLaMA2-7B-Chat2024.06 | 20.4 | 49 | |
| VanillaTarget Model=LLaMA2-7B-Chat2024.06 | 15.4 | 37.4 | |
| VecAddTarget Model=LLaMA2-13B-Chat2024.06 | 14.9 | 68.9 | |
| VecAddTarget Model=LLaMA2-7B-Chat2024.06 | 10.3 | 52.2 |