Factual Precision Evaluation on Bios
83FACTSCOREMistralINST
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MistralINSTScenario=INFO, CORE=w/o2024.07 | 83 | 75.9 | 84.8 | |
| GPT-2Scenario=INFO, CORE=w/o2024.07 | 82.2 | 78.1 | 70.3 | |
| MistralINSTScenario=REP, CORE=w/o2024.07 | 78 | 78.1 | 80.6 | |
| MistralINSTScenario=NORMAL, CORE=w/o2024.07 | 54 | 53.9 | 61.7 | |
| MistralINSTScenario=NORMAL, CORE=w/2024.07 | 49.6 | 48.3 | 61.3 | |
| MistralINSTScenario=INFO, CORE=w/2024.07 | 36.2 | 43.6 | 29.6 | |
| GPT-2Scenario=REP, CORE=w/o2024.07 | 35.4 | 40.5 | 36 | |
| MistralINSTScenario=REP, CORE=w/2024.07 | 21.9 | 26 | 14.5 | |
| GPT-2Scenario=REP, CORE=w/2024.07 | 5.35 | 7.32 | 4.37 | |
| GPT-2Scenario=INFO, CORE=w/2024.07 | 0.68 | 2.16 | 0.35 |