List your company

Targeted Attack Detection on Alpaca OnlyTarget Short

100TPRPPL
-4235077Mar 2, 2026

Evaluation results

MethodLinks
PPLModel=Qwen2.5 14B, Attacks=Ignore2026.031000.3008
PPLModel=Qwen2.5 14B, Attacks=Tree2026.031000.3008
CleanGenModel=Qwen2.5 7B, Attacks=Word2026.031000.776
CleanGenModel=Qwen2.5 7B, Attacks=Phrase2026.031000.74
CleanGenModel=Qwen2.5 7B, Attacks=Sentence2026.031000.768
STRIPModel=Qwen2.5 7B, Attacks=Joint2026.031000.796
CleanGenModel=Qwen2.5 7B, Attacks=Joint2026.031000.768
DualSentinelModel=Qwen2.5 14B, Attacks=Complete2026.0399.60.03
PPLModel=Qwen2.5 7B, Attacks=Joint2026.0399.190.344
PPLModel=Qwen2.5 7B, Attacks=Phrase2026.0399.170.344
DualSentinelModel=Qwen2.5 7B, Attacks=Word2026.0398.790.0564
STRIPModel=Qwen2.5 7B, Attacks=Word2026.0398.390.748
PPLModel=Qwen2.5 14B, Attacks=Complete2026.0398.380.3008
DualSentinelModel=Qwen2.5 7B, Attacks=Joint2026.0398.190.057
STRIPModel=Qwen2.5 14B, Attacks=Complete2026.0396.080.624
STRIPModel=Qwen2.5 14B, Attacks=Tree2026.0395.230.624
STRIPModel=Qwen2.5 7B, Attacks=Sentence2026.0394.890.804
DualSentinelModel=Qwen2.5 14B, Attacks=System2026.0392.60.03
STRIPModel=Qwen2.5 7B, Attacks=Phrase2026.0392.560.812
PPLModel=Qwen2.5 7B, Attacks=Word2026.0392.360.344
DualSentinelModel=Qwen2.5 14B, Attacks=Ignore2026.03910.03
DualSentinelModel=Qwen2.5 7B, Attacks=Sentence2026.0389.610.062
DualSentinelModel=Qwen2.5 14B, Attacks=Tree2026.0388.60.03
CleanGenModel=Qwen2.5 14B, Attacks=System2026.0387.670.7355
STRIPModel=Qwen2.5 14B, Attacks=Ignore2026.0385.60.624
DualSentinelModel=Qwen2.5 7B, Attacks=Phrase2026.0379.850.058
PPLModel=Qwen2.5 14B, Attacks=System2026.0377.730.3008
PPLModel=Qwen2.5 7B, Attacks=Sentence2026.0376.590.344
STRIPModel=Qwen2.5 14B, Attacks=System2026.0375.550.624
ONIONModel=Qwen2.5 7B, Attacks=Word2026.0365.860.008
CleanGenModel=Qwen2.5 14B, Attacks=Ignore2026.0363.920.7355
CleanGenModel=Qwen2.5 14B, Attacks=Tree2026.0361.90.7355
ONIONModel=Qwen2.5 7B, Attacks=Joint2026.0351.40.008
ParagraphModel=Qwen2.5 7B, Attacks=Word2026.0348.190.504
ParagraphModel=Qwen2.5 7B, Attacks=Sentence2026.0337.440.492
ParagraphModel=Qwen2.5 7B, Attacks=Phrase2026.0332.640.4
CleanGenModel=Qwen2.5 14B, Attacks=Complete2026.0313.760.7355
ParagraphModel=Qwen2.5 14B, Attacks=Complete2026.037.70.2896
ParagraphModel=Qwen2.5 14B, Attacks=Ignore2026.036.860.2896
ParagraphModel=Qwen2.5 14B, Attacks=System2026.034.90.2896
ParagraphModel=Qwen2.5 7B, Attacks=Joint2026.034.010.408
ONIONModel=Qwen2.5 7B, Attacks=Phrase2026.033.710.008
ParagraphModel=Qwen2.5 14B, Attacks=Tree2026.033.170.2896
ONIONModel=Qwen2.5 14B, Attacks=Complete2026.031.70.01
ONIONModel=Qwen2.5 7B, Attacks=Sentence2026.030.850.008
ONIONModel=Qwen2.5 14B, Attacks=Ignore2026.030.20.01
ConfGuardModel=Qwen2.5 14B, Attacks=Ignore2026.030.20.15
ConfGuardModel=Qwen2.5 14B, Attacks=Tree2026.030.20.15
ONIONModel=Qwen2.5 14B, Attacks=System2026.030.10.01
ConfGuardModel=Qwen2.5 14B, Attacks=Complete2026.0300.15
ONIONModel=Qwen2.5 14B, Attacks=Tree2026.0300.01
ConfGuardModel=Qwen2.5 14B, Attacks=System2026.0300.15
ConfGuardModel=Qwen2.5 7B, Attacks=Word2026.0300.1781
ConfGuardModel=Qwen2.5 7B, Attacks=Phrase2026.0300.2
ConfGuardModel=Qwen2.5 7B, Attacks=Sentence2026.0300.197
ConfGuardModel=Qwen2.5 7B, Attacks=Joint2026.0300.187