List your company

Detection Efficiency on Alpaca OnlyTarget Long (malicious)

8.378ATGRParagraph
0.509362.552184.5956.63782Mar 2, 2026

Evaluation results

MethodLinks
ParagraphModel=Qwen2.5 14B, Attack=Ignore2026.038.378
ParagraphModel=Qwen2.5 14B, Attack=System2026.038.016
ParagraphModel=Qwen2.5 14B, Attack=Tree2026.037.568
ParagraphModel=Qwen2.5 14B, Attack=Complete2026.036.998
ParagraphModel=Qwen2.5 7B, Attack=Word2026.036.743
ParagraphModel=Qwen2.5 7B, Attack=Phrase2026.036
ParagraphModel=Qwen2.5 7B, Attack=Joint2026.035.526
STRIPModel=Qwen2.5 14B, Attack=Tree2026.035.353
ParagraphModel=Qwen2.5 7B, Attack=Sentence2026.035.181
STRIPModel=Qwen2.5 14B, Attack=System2026.034.892
STRIPModel=Qwen2.5 14B, Attack=Ignore2026.033.836
STRIPModel=Qwen2.5 14B, Attack=Complete2026.033.724
ONIONModel=Qwen2.5 14B, Attack=Tree2026.033.555
ONIONModel=Qwen2.5 14B, Attack=Complete2026.033.466
STRIPModel=Qwen2.5 7B, Attack=Word2026.033.194
STRIPModel=Qwen2.5 7B, Attack=Joint2026.032.912
STRIPModel=Qwen2.5 7B, Attack=Phrase2026.032.895
STRIPModel=Qwen2.5 7B, Attack=Sentence2026.032.864
ONIONModel=Qwen2.5 7B, Attack=Phrase2026.032.386
ONIONModel=Qwen2.5 14B, Attack=Ignore2026.032.303
ONIONModel=Qwen2.5 7B, Attack=Word2026.032.108
ONIONModel=Qwen2.5 7B, Attack=Joint2026.032.09
ONIONModel=Qwen2.5 14B, Attack=System2026.031.649
ONIONModel=Qwen2.5 7B, Attack=Sentence2026.031.642
CleanGenModel=Qwen2.5 14B, Attack=System2026.031.082
CleanGenModel=Qwen2.5 14B, Attack=Complete2026.031.079
CleanGenModel=Qwen2.5 14B, Attack=Tree2026.031.074
PPLModel=Qwen2.5 7B, Attack=Sentence2026.031.071
PPLModel=Qwen2.5 14B, Attack=System2026.031.069
PPLModel=Qwen2.5 14B, Attack=Ignore2026.031.066
CleanGenModel=Qwen2.5 14B, Attack=Ignore2026.031.061
CleanGenModel=Qwen2.5 7B, Attack=Word2026.031.054
DualSentinelModel=Qwen2.5 14B, Attack=System2026.031.052
CleanGenModel=Qwen2.5 7B, Attack=Sentence2026.031.045
ConfGuardModel=Qwen2.5 7B, Attack=Joint2026.031.041
PPLModel=Qwen2.5 7B, Attack=Word2026.031.038
PPLModel=Qwen2.5 14B, Attack=Tree2026.031.029
DualSentinelModel=Qwen2.5 14B, Attack=Tree2026.031.029
PPLModel=Qwen2.5 14B, Attack=Complete2026.031.028
CleanGenModel=Qwen2.5 7B, Attack=Joint2026.031.021
ConfGuardModel=Qwen2.5 7B, Attack=Word2026.031.019
ConfGuardModel=Qwen2.5 7B, Attack=Sentence2026.031.019
ConfGuardModel=Qwen2.5 14B, Attack=System2026.031.01
DualSentinelModel=Qwen2.5 7B, Attack=Sentence2026.031.009
ConfGuardModel=Qwen2.5 14B, Attack=Ignore2026.031.007
ConfGuardModel=Qwen2.5 14B, Attack=Complete2026.031.005
DualSentinelModel=Qwen2.5 14B, Attack=Complete2026.031.003
CleanGenModel=Qwen2.5 7B, Attack=Phrase2026.031.003
PPLModel=Qwen2.5 7B, Attack=Joint2026.031.001
ConfGuardModel=Qwen2.5 7B, Attack=Phrase2026.030.982
DualSentinelModel=Qwen2.5 7B, Attack=Joint2026.030.972
DualSentinelModel=Qwen2.5 7B, Attack=Word2026.030.956
PPLModel=Qwen2.5 7B, Attack=Phrase2026.030.951
DualSentinelModel=Qwen2.5 7B, Attack=Phrase2026.030.944
ConfGuardModel=Qwen2.5 14B, Attack=Tree2026.030.902
DualSentinelModel=Qwen2.5 14B, Attack=Ignore2026.030.812