Factual Question Answering on LSQA OOD
79.56PrecisionUALIGN
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| UALIGNBackbone=Llama-3-8B2024.12 | 79.56 | 55.88 | |
| ICL-COTBackbone=Mistral-7B2024.12 | 79.24 | 52.59 | |
| ICLBackbone=Llama-3-8B2024.12 | 77.35 | 52.98 | |
| UALIGNBackbone=Mistral-7B2024.12 | 76.29 | 52.89 | |
| UALIGN (w/o Entro.)Backbone=Llama-3-8B2024.12 | 76.14 | 54.43 | |
| UALIGN (w/o Entro.)Backbone=Mistral-7B2024.12 | 75.34 | 50.15 | |
| RLKFBackbone=Llama-3-8B2024.12 | 74.95 | 52.46 | |
| RL-PPOBackbone=Mistral-7B2024.12 | 74.49 | 49.67 | |
| UALIGN (w/o Conf.)Backbone=Llama-3-8B2024.12 | 74.25 | 52.06 | |
| RLKFBackbone=Mistral-7B2024.12 | 74.19 | 49.23 | |
| ICLBackbone=Mistral-7B2024.12 | 74.09 | 47.71 | |
| ICL-COTBackbone=Llama-3-8B2024.12 | 73.96 | 49.37 | |
| UALIGN (w/o Conf.)Backbone=Mistral-7B2024.12 | 73.15 | 47.06 | |
| ICL-IDKBackbone=Mistral-7B2024.12 | 72.27 | 47.32 | |
| RL-PPOBackbone=Llama-3-8B2024.12 | 72.18 | 48.43 | |
| R-TuningBackbone=Llama-3-8B2024.12 | 71.54 | 52.15 | |
| RL-DPOBackbone=Llama-3-8B2024.12 | 71.09 | 48.88 | |
| ITIBackbone=Llama-3-8B2024.12 | 68.52 | 46.99 | |
| SFTBackbone=Mistral-7B2024.12 | 68.33 | 44 | |
| SFTBackbone=Llama-3-8B2024.12 | 68.09 | 46.63 | |
| RL-DPOBackbone=Mistral-7B2024.12 | 67.82 | 43.77 | |
| ITIBackbone=Mistral-7B2024.12 | 67.73 | 46.2 | |
| ICL-IDKBackbone=Llama-3-8B2024.12 | 66.67 | 50.24 | |
| R-TuningBackbone=Mistral-7B2024.12 | 64.67 | 44.05 |