Safety Evaluation on CValues
89.25AccuracyROSE
Evaluation Results
| Method | Links | |
|---|---|---|
| ROSEModel=Qwen-chat-7B, Decoding=ROSE2024.02 | 89.25 | |
| Qwen-chat-7BModel=Qwen-chat-7B, Decoding=Regular2024.02 | 89.19 | |
| ROSEModel=InternLM-chat-7B, Decoding=ROSE2024.02 | 85.92 | |
| InternLM-chat-7BModel=InternLM-chat-7B, Decoding=Regular2024.02 | 85.28 | |
| ROSEModel=Chinese-Alpaca-7B, Decoding=ROSE2024.02 | 84.22 | |
| Chinese-Alpaca-7BModel=Chinese-Alpaca-7B, Decoding=Regular2024.02 | 80.37 | |
| ROSEModel=Alpaca-7B, Decoding=ROSE2024.02 | 72.49 | |
| Alpaca-7BModel=Alpaca-7B, Decoding=Regular2024.02 | 68.81 | |
| ROSEModel=Vicuna-7B, Decoding=ROSE2024.02 | 67.82 | |
| Vicuna-7BModel=Vicuna-7B, Decoding=Regular2024.02 | 64.89 |