Factuality Evaluation on MMLU
82.4EMGPT-4
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-4Type=Proprietary, Shots=0-shot2023.06 | 82.4 | |
| ChatGPTType=Proprietary, Shots=0-shot2023.06 | 67.9 | |
| ShareGPT 65BBase Model=LLaMA 65B, Training Data=ShareGPT, Shots=0-shot2023.06 | 61.3 | |
| Human mix. 65BBase Model=LLaMA 65B, Training Data=Human mix., Shots=0-shot2023.06 | 60.4 | |
| TÜLU 65BBase Model=LLaMA, Training Data=Human+GPT, Shots=0-shot2023.06 | 59.2 | |
| LLaMa 65BBase Model=LLaMA, Shots=0-shot2023.06 | 58.7 | |
| TÜLU 30BBase Model=LLaMA, Training Data=Human+GPT, Shots=0-shot2023.06 | 57.7 | |
| LLaMa 30BBase Model=LLaMA, Shots=0-shot2023.06 | 54.6 | |
| TÜLU-1.1 13BBase Model=LLaMA-2, Training Data=Human+GPT, Shots=0-shot2023.06 | 52.3 | |
| LLaMa-2 13BBase Model=LLaMA-2, Shots=0-shot2023.06 | 52 | |
| TÜLU 13BBase Model=LLaMA, Training Data=Human+GPT, Shots=0-shot2023.06 | 49.3 | |
| TÜLU-1.1 7BBase Model=LLaMA-2, Training Data=Human+GPT, Shots=0-shot2023.06 | 49.2 | |
| TÜLU 7BBase Model=LLaMA, Training Data=Human+GPT, Shots=0-shot2023.06 | 44.8 | |
| LLaMa 13BBase Model=LLaMA, Shots=0-shot2023.06 | 42.3 | |
| LLaMa-2 7BBase Model=LLaMA-2, Shots=0-shot2023.06 | 41.8 | |
| LLaMa 7BBase Model=LLaMA, Shots=0-shot2023.06 | 31.5 |