Factual Question Answering on NQ-Open ID
57.34PrecisionICL-COT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ICL-COTBackbone=Llama-3-8B2024.12 | 57.34 | 23.6 | |
| UALIGN (w/o Entro.)Backbone=Llama-3-8B2024.12 | 56.68 | 27.56 | |
| UALIGNBackbone=Llama-3-8B2024.12 | 56.6 | 26.09 | |
| RLKFBackbone=Llama-3-8B2024.12 | 54.86 | 22.15 | |
| ICL-COTBackbone=Mistral-7B2024.12 | 54.47 | 18.22 | |
| UALIGN (w/o Conf.)Backbone=Llama-3-8B2024.12 | 54.19 | 23.6 | |
| UALIGNBackbone=Mistral-7B2024.12 | 54.17 | 19.64 | |
| RL-PPOBackbone=Llama-3-8B2024.12 | 54.07 | 24.19 | |
| RL-PPOBackbone=Mistral-7B2024.12 | 53.76 | 19.19 | |
| UALIGN (w/o Entro.)Backbone=Mistral-7B2024.12 | 52.21 | 18.44 | |
| ICLBackbone=Mistral-7B2024.12 | 52.09 | 17.95 | |
| RLKFBackbone=Mistral-7B2024.12 | 52.09 | 18.24 | |
| ICL-IDKBackbone=Mistral-7B2024.12 | 51.13 | 17.67 | |
| ITIBackbone=Mistral-7B2024.12 | 51.12 | 16.68 | |
| RL-DPOBackbone=Mistral-7B2024.12 | 50.95 | 16.42 | |
| SFTBackbone=Mistral-7B2024.12 | 50.82 | 14.42 | |
| UALIGN (w/o Conf.)Backbone=Mistral-7B2024.12 | 50.3 | 17.88 | |
| ICLBackbone=Llama-3-8B2024.12 | 50.28 | 20.11 | |
| RL-DPOBackbone=Llama-3-8B2024.12 | 49.65 | 19.18 | |
| R-TuningBackbone=Llama-3-8B2024.12 | 47.81 | 18.12 | |
| R-TuningBackbone=Mistral-7B2024.12 | 46.33 | 15.52 | |
| ICL-IDKBackbone=Llama-3-8B2024.12 | 45.43 | 20.72 | |
| ITIBackbone=Llama-3-8B2024.12 | 43.2 | 17.13 | |
| SFTBackbone=Llama-3-8B2024.12 | 41.41 | 16.57 |