Instruction Following on AlpacaEval v1 (test)
97.7AlpacaEval ScoreGPT-4-Turbo
Evaluation results
| Method | Links | |
|---|---|---|
| GPT-4-Turbo2024.03 | 97.7 | |
| Yi-ChatSize=34B2024.03 | 94.08 | |
| LLaMA2-ChatSize=70B2024.03 | 92.66 | |
| GPT-3.5-Turbo2024.03 | 89.37 | |
| RECOSTData size=5200 (10%), Training Dataset=Alpaca-gpt42024.02 | 63.35 | |
| Full Alpaca-gpt4Data size=52002 (100%), Training Dataset=Alpaca-gpt42024.02 | 61.8 | |
| Predictive EntropyData size=5200 (10%), Training Dataset=Alpaca-gpt42024.02 | 60.39 | |
| RandomData size=5200 (10%), Training Dataset=Alpaca-gpt42024.02 | 55.92 | |
| RECOSTData size=520 (1%), Training Dataset=Alpaca2024.02 | 39.19 | |
| IFDData size=3111 (6%), Training Dataset=Alpaca2024.02 | 36.78 | |
| Predictive EntropyData size=520 (1%), Training Dataset=Alpaca2024.02 | 35.59 | |
| Full AlpacaData size=52002 (100%), Training Dataset=Alpaca2024.02 | 27.75 | |
| LIMAData size=10002024.02 | 26.58 | |
| RandomData size=520 (1%), Training Dataset=Alpaca2024.02 | 26.52 |