Excessive Response Attack on SecLens Human Verified (test)
53.37Attack Success RateSecLens
Evaluation Results
| Method | Links | |
|---|---|---|
| SecLensModel=GPT-4-turbo2025.06 | 53.37 | |
| SecLensModel=GPT-4o2025.06 | 52.89 | |
| MCTSModel=GPT-4-turbo2025.06 | 49.12 | |
| SecLensModel=Claude-3.72025.06 | 47.85 | |
| MCTSModel=GPT-4o2025.06 | 47.84 | |
| MCTSModel=Claude-3.72025.06 | 44.31 | |
| TuningModel=GPT-4-turbo2025.06 | 39.65 | |
| TuningModel=GPT-4o2025.06 | 38.91 | |
| TuningModel=Claude-3.72025.06 | 35.27 | |
| RandomModel=GPT-4-turbo2025.06 | 14.55 | |
| RandomModel=GPT-4o2025.06 | 13.72 | |
| RandomModel=Claude-3.72025.06 | 9.83 |