Secondary-risk discovery on Excessive Response
74.64Attack Success Rate (ASR)SecLens
Evaluation results
| Method | Links | |
|---|---|---|
| SecLensTarget Model=Gemma-2-27b2025.06 | 74.64 | |
| SecLensTarget Model=Qwen2.5-32b2025.06 | 72.84 | |
| SecLensTarget Model=Deepseek-v32025.06 | 72.09 | |
| SecLensTarget Model=GPT-4-turbo2025.06 | 71.52 | |
| SecLensTarget Model=Gemini 2.0-pro2025.06 | 70.55 | |
| SecLensTarget Model=Llama-3.3-70b2025.06 | 69.41 | |
| PIFTarget Model=Deepseek-v32025.06 | 69.22 | |
| SecLensTarget Model=GPT-4o2025.06 | 68.74 | |
| PIFTarget Model=Gemma-2-27b2025.06 | 68.11 | |
| AutoBreachTarget Model=Deepseek-v32025.06 | 67.84 | |
| PIFTarget Model=Phi-42025.06 | 66.91 | |
| AutoBreachTarget Model=Gemma-2-27b2025.06 | 66.72 | |
| PIFTarget Model=GPT-4o2025.06 | 66.48 | |
| PIFTarget Model=Gemini 2.0-pro2025.06 | 66.42 | |
| PIFTarget Model=Qwen2.5-32b2025.06 | 66.31 | |
| SecLensTarget Model=Phi-42025.06 | 65.48 | |
| PIFTarget Model=GPT-4-turbo2025.06 | 64.74 | |
| AutoBreachTarget Model=GPT-4o2025.06 | 64.12 | |
| AutoBreachTarget Model=Phi-42025.06 | 63.88 | |
| AutoBreachTarget Model=Qwen2.5-32b2025.06 | 63.27 | |
| AutoBreachTarget Model=GPT-4-turbo2025.06 | 62.9 | |
| PIFTarget Model=Llama-3.3-70b2025.06 | 62.14 | |
| AutoBreachTarget Model=Gemini 2.0-pro2025.06 | 62.05 | |
| AutoBreachTarget Model=Llama-3.3-70b2025.06 | 60.95 | |
| SecLensTarget Model=Claude-3.72025.06 | 59.71 | |
| PIFTarget Model=Claude-3.72025.06 | 54.13 | |
| AutoBreachTarget Model=Claude-3.72025.06 | 53.75 |