Factual Question Answering on ID Datasets Average
70.82PrecisionUALIGN
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| UALIGNBackbone=Mistral-7B2024.12 | 70.82 | 41.79 | |
| UALIGNBackbone=Llama-3-8B2024.12 | 70.72 | 43.71 | |
| UALIGN (w/o Entro.)Backbone=Llama-3-8B2024.12 | 70.16 | 44.25 | |
| UALIGN (w/o Entro.)Backbone=Mistral-7B2024.12 | 68.43 | 40.56 | |
| RLKFBackbone=Llama-3-8B2024.12 | 68.11 | 41.04 | |
| RL-PPOBackbone=Mistral-7B2024.12 | 68.11 | 39.45 | |
| RL-PPOBackbone=Llama-3-8B2024.12 | 68.03 | 41.72 | |
| RLKFBackbone=Mistral-7B2024.12 | 67.72 | 39.68 | |
| ICL-COTBackbone=Mistral-7B2024.12 | 67.69 | 39.06 | |
| UALIGN (w/o Conf.)Backbone=Llama-3-8B2024.12 | 67.45 | 41.01 | |
| ICLBackbone=Mistral-7B2024.12 | 66.21 | 38.43 | |
| ICLBackbone=Llama-3-8B2024.12 | 65.62 | 40.32 | |
| UALIGN (w/o Conf.)Backbone=Mistral-7B2024.12 | 65.52 | 38.83 | |
| ICL-COTBackbone=Llama-3-8B2024.12 | 65.45 | 40.95 | |
| RL-DPOBackbone=Llama-3-8B2024.12 | 64.32 | 39.11 | |
| ITIBackbone=Mistral-7B2024.12 | 64.22 | 38.91 | |
| R-TuningBackbone=Llama-3-8B2024.12 | 64.04 | 39.48 | |
| SFTBackbone=Mistral-7B2024.12 | 63.74 | 37.08 | |
| RL-DPOBackbone=Mistral-7B2024.12 | 63.19 | 37.06 | |
| ICL-IDKBackbone=Mistral-7B2024.12 | 62.48 | 36.2 | |
| ITIBackbone=Llama-3-8B2024.12 | 62.21 | 38.3 | |
| SFTBackbone=Llama-3-8B2024.12 | 61.46 | 38.18 | |
| ICL-IDKBackbone=Llama-3-8B2024.12 | 60.98 | 39.27 | |
| R-TuningBackbone=Mistral-7B2024.12 | 59.49 | 36.29 |