List your company

Targeted Attack Detection on Alpaca AddTarget Medium

100TPRConfGuard
-3.58423.30850.277.092Mar 2, 2026

Evaluation results

MethodLinks
ConfGuardModel=Qwen2.5 7B, Attacks=Phrase2026.0310027.01
ConfGuardModel=Qwen2.5 7B, Attacks=Word2026.0399.8826.16
ConfGuardModel=Qwen2.5 14B, Attacks=System2026.0399.615
PPLModel=Qwen2.5 7B, Attacks=Joint2026.0399.5834.67
ConfGuardModel=Qwen2.5 7B, Attacks=Joint2026.0399.5827.41
DualSentinelModel=Qwen2.5 7B, Attacks=Joint2026.0399.5812.95
PPLModel=Qwen2.5 7B, Attacks=Phrase2026.0399.5434.67
DualSentinelModel=Qwen2.5 7B, Attacks=Phrase2026.0399.548.5
PPLModel=Qwen2.5 14B, Attacks=System2026.0399.530.08
DualSentinelModel=Qwen2.5 7B, Attacks=Word2026.0399.2110.23
DualSentinelModel=Qwen2.5 14B, Attacks=System2026.0399.23
ConfGuardModel=Qwen2.5 7B, Attacks=Sentence2026.0398.6628.34
STRIPModel=Qwen2.5 7B, Attacks=Joint2026.0398.3288.3
STRIPModel=Qwen2.5 7B, Attacks=Word2026.0397.7585.82
STRIPModel=Qwen2.5 7B, Attacks=Sentence2026.0397.3388.25
CleanGenModel=Qwen2.5 7B, Attacks=Sentence2026.0397.3378.13
DualSentinelModel=Qwen2.5 7B, Attacks=Sentence2026.0397.3312.6
STRIPModel=Qwen2.5 7B, Attacks=Phrase2026.0397.2685.88
CleanGenModel=Qwen2.5 7B, Attacks=Word2026.0394.1776.92
CleanGenModel=Qwen2.5 7B, Attacks=Phrase2026.0393.1579.43
PPLModel=Qwen2.5 7B, Attacks=Word2026.0392.8234.41
CleanGenModel=Qwen2.5 7B, Attacks=Joint2026.0392.0578.62
CleanGenModel=Qwen2.5 14B, Attacks=System2026.038773.55
PPLModel=Qwen2.5 7B, Attacks=Sentence2026.038033.6
ParagraphModel=Qwen2.5 7B, Attacks=Joint2026.0374.0551.2
ONIONModel=Qwen2.5 7B, Attacks=Word2026.0363.220.4
ParagraphModel=Qwen2.5 14B, Attacks=System2026.0359.228.96
STRIPModel=Qwen2.5 14B, Attacks=System2026.0356.862.4
ONIONModel=Qwen2.5 7B, Attacks=Joint2026.0345.60.8
ParagraphModel=Qwen2.5 7B, Attacks=Sentence2026.0322.6648.17
ParagraphModel=Qwen2.5 7B, Attacks=Phrase2026.0320.0944.35
ParagraphModel=Qwen2.5 7B, Attacks=Word2026.0315.6939.67
ONIONModel=Qwen2.5 7B, Attacks=Phrase2026.033.650.8
ONIONModel=Qwen2.5 7B, Attacks=Sentence2026.031.330.4
ONIONModel=Qwen2.5 14B, Attacks=System2026.030.41