List your company

Detection Efficiency on Alpaca OnlyTarget Long (benign)

6.115ATGRParagraph
0.797482.177993.55854.93901Mar 2, 2026

Evaluation results

MethodLinks
ParagraphModel=Qwen2.5 14B, Attack=Ignore2026.036.115
ParagraphModel=Qwen2.5 14B, Attack=Complete2026.036.115
ParagraphModel=Qwen2.5 14B, Attack=Tree2026.036.115
ParagraphModel=Qwen2.5 14B, Attack=System2026.036.115
ParagraphModel=Qwen2.5 7B, Attack=Word2026.034.666
ParagraphModel=Qwen2.5 7B, Attack=Sentence2026.034.422
ParagraphModel=Qwen2.5 7B, Attack=Phrase2026.034.342
ParagraphModel=Qwen2.5 7B, Attack=Joint2026.034.238
STRIPModel=Qwen2.5 14B, Attack=Ignore2026.033.778
STRIPModel=Qwen2.5 14B, Attack=Complete2026.033.778
STRIPModel=Qwen2.5 14B, Attack=Tree2026.033.778
STRIPModel=Qwen2.5 14B, Attack=System2026.033.778
STRIPModel=Qwen2.5 7B, Attack=Word2026.032.666
STRIPModel=Qwen2.5 7B, Attack=Sentence2026.032.555
STRIPModel=Qwen2.5 7B, Attack=Phrase2026.032.509
STRIPModel=Qwen2.5 7B, Attack=Joint2026.032.413
ONIONModel=Qwen2.5 14B, Attack=Ignore2026.031.306
ONIONModel=Qwen2.5 14B, Attack=Complete2026.031.306
ONIONModel=Qwen2.5 14B, Attack=Tree2026.031.306
ONIONModel=Qwen2.5 14B, Attack=System2026.031.306
ONIONModel=Qwen2.5 7B, Attack=Phrase2026.031.188
ONIONModel=Qwen2.5 7B, Attack=Sentence2026.031.185
ONIONModel=Qwen2.5 7B, Attack=Joint2026.031.148
ONIONModel=Qwen2.5 7B, Attack=Word2026.031.133
PPLModel=Qwen2.5 7B, Attack=Sentence2026.031.067
ConfGuardModel=Qwen2.5 7B, Attack=Joint2026.031.051
CleanGenModel=Qwen2.5 14B, Attack=Ignore2026.031.039
CleanGenModel=Qwen2.5 14B, Attack=Complete2026.031.039
CleanGenModel=Qwen2.5 14B, Attack=Tree2026.031.039
CleanGenModel=Qwen2.5 14B, Attack=System2026.031.039
CleanGenModel=Qwen2.5 7B, Attack=Joint2026.031.039
CleanGenModel=Qwen2.5 7B, Attack=Sentence2026.031.034
PPLModel=Qwen2.5 7B, Attack=Phrase2026.031.027
ConfGuardModel=Qwen2.5 7B, Attack=Sentence2026.031.023
PPLModel=Qwen2.5 7B, Attack=Word2026.031.021
PPLModel=Qwen2.5 14B, Attack=Ignore2026.031.019
PPLModel=Qwen2.5 14B, Attack=Complete2026.031.019
PPLModel=Qwen2.5 14B, Attack=Tree2026.031.019
PPLModel=Qwen2.5 14B, Attack=System2026.031.019
CleanGenModel=Qwen2.5 7B, Attack=Word2026.031.019
DualSentinelModel=Qwen2.5 7B, Attack=Sentence2026.031.018
DualSentinelModel=Qwen2.5 7B, Attack=Joint2026.031.013
ConfGuardModel=Qwen2.5 7B, Attack=Word2026.031.01
ConfGuardModel=Qwen2.5 7B, Attack=Phrase2026.031.008
DualSentinelModel=Qwen2.5 14B, Attack=Ignore2026.031.007
DualSentinelModel=Qwen2.5 14B, Attack=Complete2026.031.007
DualSentinelModel=Qwen2.5 14B, Attack=Tree2026.031.007
DualSentinelModel=Qwen2.5 14B, Attack=System2026.031.007
ConfGuardModel=Qwen2.5 14B, Attack=Ignore2026.031.006
ConfGuardModel=Qwen2.5 14B, Attack=Complete2026.031.006
ConfGuardModel=Qwen2.5 14B, Attack=Tree2026.031.006
ConfGuardModel=Qwen2.5 14B, Attack=System2026.031.006
DualSentinelModel=Qwen2.5 7B, Attack=Phrase2026.031.006
DualSentinelModel=Qwen2.5 7B, Attack=Word2026.031.003
CleanGenModel=Qwen2.5 7B, Attack=Phrase2026.031.003
PPLModel=Qwen2.5 7B, Attack=Joint2026.031.002