Question Answering on TruthfulQA (Accuracy, Precision, Recall, F1)
52.26AccuracyMistral-7B-instruct
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Mistral-7B-instruct2025.06 | 52.26 | — | — | — | |
| KD (gre-box)Base Model=Llama3-8B, Strategy=Grey-box Knowledge Distillation2024.09 | 46.3 | 50 | 23.2 | 31.6 | |
| PUGC+DPO2025.06 | 42.77 | — | — | — | |
| Victim Model (GPT-3.5-turbo)Model Type=Victim2024.09 | 41.4 | 50 | 20.7 | 29.3 | |
| LoRDBase Model=Llama3-8B, Strategy=Locality Reinforced Distillation2024.09 | 40.8 | 50 | 20.4 | 28.9 | |
| Local Model (Llama3-8B)Model Type=Local initial model2024.09 | 39.1 | 50 | 19.5 | 28.1 | |
| MLEBase Model=Llama3-8B, Strategy=Model Extraction (MLE)2024.09 | 38.1 | 50 | 19 | 26.6 |