Factuality Evaluation on NQ-Swap
43.7Science Category ScoreCoDA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CoDABackbone=Mistral-7b2025.02 | 43.7 | — | 27.7 | |
| CoTBackbone=Mistral-7b2025.02 | 39 | — | 19.5 | |
| CoDABackbone=Llama-2-7b-chat2025.02 | 38.9 | — | 26.8 | |
| DolaBackbone=Mistral-7b2025.02 | 38.4 | — | 15.9 | |
| SRBackbone=Mistral-7b2025.02 | 38.2 | — | 13.8 | |
| CoTBackbone=Llama-2-7b-chat2025.02 | 36.7 | — | 19.2 | |
| GreedyBackbone=Mistral-7b2025.02 | 36.7 | — | 12.6 | |
| USCBackbone=Mistral-7b2025.02 | 35.9 | — | 11.4 | |
| SRBackbone=Llama-2-7b-chat2025.02 | 35.8 | — | 14.2 | |
| GreedyBackbone=Llama-2-7b-chat2025.02 | 33.4 | — | 8.5 | |
| DolaBackbone=Llama-2-7b-chat2025.02 | 33 | — | 13.8 | |
| USCBackbone=Llama-2-7b-chat2025.02 | 32.9 | — | 9.4 |