Factual Question Answering on SciQ
76.44PrecisionUALIGN
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| UALIGNBackbone=Llama-3-8B2024.12 | 76.44 | 48 | |
| RL-PPOBackbone=Llama-3-8B2024.12 | 75.7 | 45.8 | |
| UALIGN (w/o Entro.)Backbone=Llama-3-8B2024.12 | 75.39 | 47.5 | |
| UALIGN (w/o Conf.)Backbone=Llama-3-8B2024.12 | 74.05 | 45 | |
| UALIGNBackbone=Mistral-7B2024.12 | 73.21 | 46.7 | |
| UALIGN (w/o Entro.)Backbone=Mistral-7B2024.12 | 72.9 | 45.6 | |
| RLKFBackbone=Llama-3-8B2024.12 | 72.36 | 44.9 | |
| ITIBackbone=Llama-3-8B2024.12 | 72.35 | 43.8 | |
| ICL-COTBackbone=Llama-3-8B2024.12 | 72.34 | 45.9 | |
| SFTBackbone=Llama-3-8B2024.12 | 72.18 | 45.4 | |
| ICL-COTBackbone=Mistral-7B2024.12 | 71.87 | 44.2 | |
| R-TuningBackbone=Llama-3-8B2024.12 | 71.38 | 44.9 | |
| RL-PPOBackbone=Mistral-7B2024.12 | 71.35 | 44.1 | |
| RL-DPOBackbone=Llama-3-8B2024.12 | 71.23 | 44.2 | |
| RLKFBackbone=Mistral-7B2024.12 | 70.66 | 43.9 | |
| ICLBackbone=Llama-3-8B2024.12 | 70.43 | 44.3 | |
| UALIGN (w/o Conf.)Backbone=Mistral-7B2024.12 | 69.84 | 43.5 | |
| ICLBackbone=Mistral-7B2024.12 | 68.62 | 42.2 | |
| ICL-IDKBackbone=Llama-3-8B2024.12 | 68.36 | 43 | |
| ITIBackbone=Mistral-7B2024.12 | 66.9 | 44.9 | |
| RL-DPOBackbone=Mistral-7B2024.12 | 66.44 | 41.8 | |
| SFTBackbone=Mistral-7B2024.12 | 65.85 | 42.5 | |
| R-TuningBackbone=Mistral-7B2024.12 | 64.44 | 40.1 | |
| ICL-IDKBackbone=Mistral-7B2024.12 | 63.74 | 39.2 |