Jailbreaking on NeuroStrike Evaluation Prompts
43.2Attack Success Rate (ASR)NeuroStrike
Evaluation Results
| Method | Links | |
|---|---|---|
| NeuroStrikeTarget Model=Gemini-2.0-Flash-Lite, Defense Method=Perplexity Filter2025.09 | 43.2 | |
| NeuroStrikeTarget Model=QwQ-32B, Defense Method=LSE2025.09 | 43.4 | |
| NeuroStrikeTarget Model=Gemini-2.0-Flash-Lite, Defense Method=SmoothLLM2025.09 | 47.3 | |
| NeuroStrikeTarget Model=Gemini-2.0-Flash, Defense Method=Perplexity Filter2025.09 | 48.7 | |
| NeuroStrikeTarget Model=Gemini-1.5-Pro, Defense Method=Perplexity Filter2025.09 | 49.7 | |
| NeuroStrikeTarget Model=Gemini-2.0-Flash, Defense Method=SmoothLLM2025.09 | 52.8 | |
| NeuroStrikeTarget Model=Gemini-1.5-Pro, Defense Method=SmoothLLM2025.09 | 53.8 | |
| NeuroStrikeTarget Model=Gemma-3-1b-it, Defense Method=LSE2025.09 | 60 | |
| NeuroStrikeTarget Model=QwQ-32B, Defense Method=SmoothLLM2025.09 | 76.4 | |
| NeuroStrikeTarget Model=Gemma-3-1b-it, Defense Method=SmoothLLM2025.09 | 78 | |
| NeuroStrikeTarget Model=QwQ-32B, Defense Method=Perplexity Filter2025.09 | 78.8 | |
| NeuroStrikeTarget Model=Gemma-3-1b-it, Defense Method=Perplexity Filter2025.09 | 79.8 |