Adversarial Attack on SafeMT ATTACK 600
71.29Attack Success RateGPT-4o
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-4oSteering Configuration=Original2025.02 | 71.29 | |
| Llama3-70bSteering Configuration=Original2025.02 | 67.58 | |
| Llama3-70bSteering Configuration=w/ steering2025.02 | 16.08 | |
| GPT-4oSteering Configuration=w/ steering2025.02 | 15.42 |