Detection Efficiency on Alpaca OnlyTarget Long (benign)
6.115ATGRParagraph
Evaluation Results
| Method | Links | |
|---|---|---|
| ParagraphModel=Qwen2.5 14B, Attack=Ignore2026.03 | 6.115 | |
| ParagraphModel=Qwen2.5 14B, Attack=Complete2026.03 | 6.115 | |
| ParagraphModel=Qwen2.5 14B, Attack=Tree2026.03 | 6.115 | |
| ParagraphModel=Qwen2.5 14B, Attack=System2026.03 | 6.115 | |
| ParagraphModel=Qwen2.5 7B, Attack=Word2026.03 | 4.666 | |
| ParagraphModel=Qwen2.5 7B, Attack=Sentence2026.03 | 4.422 | |
| ParagraphModel=Qwen2.5 7B, Attack=Phrase2026.03 | 4.342 | |
| ParagraphModel=Qwen2.5 7B, Attack=Joint2026.03 | 4.238 | |
| STRIPModel=Qwen2.5 14B, Attack=Ignore2026.03 | 3.778 | |
| STRIPModel=Qwen2.5 14B, Attack=Complete2026.03 | 3.778 | |
| STRIPModel=Qwen2.5 14B, Attack=Tree2026.03 | 3.778 | |
| STRIPModel=Qwen2.5 14B, Attack=System2026.03 | 3.778 | |
| STRIPModel=Qwen2.5 7B, Attack=Word2026.03 | 2.666 | |
| STRIPModel=Qwen2.5 7B, Attack=Sentence2026.03 | 2.555 | |
| STRIPModel=Qwen2.5 7B, Attack=Phrase2026.03 | 2.509 | |
| STRIPModel=Qwen2.5 7B, Attack=Joint2026.03 | 2.413 | |
| ONIONModel=Qwen2.5 14B, Attack=Ignore2026.03 | 1.306 | |
| ONIONModel=Qwen2.5 14B, Attack=Complete2026.03 | 1.306 | |
| ONIONModel=Qwen2.5 14B, Attack=Tree2026.03 | 1.306 | |
| ONIONModel=Qwen2.5 14B, Attack=System2026.03 | 1.306 | |
| ONIONModel=Qwen2.5 7B, Attack=Phrase2026.03 | 1.188 | |
| ONIONModel=Qwen2.5 7B, Attack=Sentence2026.03 | 1.185 | |
| ONIONModel=Qwen2.5 7B, Attack=Joint2026.03 | 1.148 | |
| ONIONModel=Qwen2.5 7B, Attack=Word2026.03 | 1.133 | |
| PPLModel=Qwen2.5 7B, Attack=Sentence2026.03 | 1.067 | |
| ConfGuardModel=Qwen2.5 7B, Attack=Joint2026.03 | 1.051 | |
| CleanGenModel=Qwen2.5 14B, Attack=Ignore2026.03 | 1.039 | |
| CleanGenModel=Qwen2.5 14B, Attack=Complete2026.03 | 1.039 | |
| CleanGenModel=Qwen2.5 14B, Attack=Tree2026.03 | 1.039 | |
| CleanGenModel=Qwen2.5 14B, Attack=System2026.03 | 1.039 | |
| CleanGenModel=Qwen2.5 7B, Attack=Joint2026.03 | 1.039 | |
| CleanGenModel=Qwen2.5 7B, Attack=Sentence2026.03 | 1.034 | |
| PPLModel=Qwen2.5 7B, Attack=Phrase2026.03 | 1.027 | |
| ConfGuardModel=Qwen2.5 7B, Attack=Sentence2026.03 | 1.023 | |
| PPLModel=Qwen2.5 7B, Attack=Word2026.03 | 1.021 | |
| PPLModel=Qwen2.5 14B, Attack=Ignore2026.03 | 1.019 | |
| PPLModel=Qwen2.5 14B, Attack=Complete2026.03 | 1.019 | |
| PPLModel=Qwen2.5 14B, Attack=Tree2026.03 | 1.019 | |
| PPLModel=Qwen2.5 14B, Attack=System2026.03 | 1.019 | |
| CleanGenModel=Qwen2.5 7B, Attack=Word2026.03 | 1.019 | |
| DualSentinelModel=Qwen2.5 7B, Attack=Sentence2026.03 | 1.018 | |
| DualSentinelModel=Qwen2.5 7B, Attack=Joint2026.03 | 1.013 | |
| ConfGuardModel=Qwen2.5 7B, Attack=Word2026.03 | 1.01 | |
| ConfGuardModel=Qwen2.5 7B, Attack=Phrase2026.03 | 1.008 | |
| DualSentinelModel=Qwen2.5 14B, Attack=Ignore2026.03 | 1.007 | |
| DualSentinelModel=Qwen2.5 14B, Attack=Complete2026.03 | 1.007 | |
| DualSentinelModel=Qwen2.5 14B, Attack=Tree2026.03 | 1.007 | |
| DualSentinelModel=Qwen2.5 14B, Attack=System2026.03 | 1.007 | |
| ConfGuardModel=Qwen2.5 14B, Attack=Ignore2026.03 | 1.006 | |
| ConfGuardModel=Qwen2.5 14B, Attack=Complete2026.03 | 1.006 | |
| ConfGuardModel=Qwen2.5 14B, Attack=Tree2026.03 | 1.006 | |
| ConfGuardModel=Qwen2.5 14B, Attack=System2026.03 | 1.006 | |
| DualSentinelModel=Qwen2.5 7B, Attack=Phrase2026.03 | 1.006 | |
| DualSentinelModel=Qwen2.5 7B, Attack=Word2026.03 | 1.003 | |
| CleanGenModel=Qwen2.5 7B, Attack=Phrase2026.03 | 1.003 | |
| PPLModel=Qwen2.5 7B, Attack=Joint2026.03 | 1.002 |