Red-Teaming (Attack Success Rate) on HARMFULQA
0.702ASRCHATGPT
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| CHATGPTPrompting Strategy=RED-EVAL2023.08 | 0.702 | — | — | — | |
| GPT-4Prompting Strategy=RED-EVAL2023.08 | 0.452 | — | — | — | |
| CHATGPTPrompting Strategy=COT2023.08 | 0.027 | — | — | — | |
| CHATGPTPrompting Strategy=STANDARD2023.08 | 0.018 | — | — | — | |
| GPT-4Prompting Strategy=COT2023.08 | 0.004 | — | — | — | |
| GPT-4Prompting Strategy=STANDARD2023.08 | 0.001 | — | — | — | |
| CHATGPT2023.08 | 0.001 | 0.257 | 0.018 | 0.728 | |
| CHATGPTPrompting Strategy=Overall2023.08 | — | 0.249 | — | — | |
| GPT-4Prompting Strategy=Overall2023.08 | — | 0.152 | — | — |