List your company

Sentiment Classification on SST-2 (Repair and Attack Metrics)

99.67Attack Generation RateVanilla
7.692431.571255.4579.3288Nov 16, 2025Dec 8, 2025Dec 31, 2025Jan 23, 2026Feb 14, 2026Mar 9, 2026Apr 1, 2026

Evaluation results

MethodLinks
VanillaVictim=BERT, Attack=BadNets2025.1199.67———92.65
VanillaVictim=BERT, Attack=Cbat2025.1195.42———90.85
VanillaVictim=BERT, Attack=HiddenKiller2025.1194.16———89.33
VanillaVictim=BERT, Attack=BGMAttack2025.1191.27———88.21
VanillaVictim=BERT, Attack=NWS2025.1190.23———90.45
VanillaVictim=Llama2-7B, Attack=NWS2025.1189.25———85.15
VanillaVictim=Llama2-7B, Attack=BadNets2025.1188.73———86.91
VanillaVictim=Llama2-7B, Attack=Cbat2025.1185.31———84.01
VanillaVictim=Llama2-7B, Attack=BGMAttack2025.1184.19———83.2
VanillaVictim=Llama2-7B, Attack=HiddenKiller2025.1183.27———81.56
WARPModel=DistilBERT2026.046910010091.4—
WARPModel=BERT2026.0464.710010092.7—
Same-layer LoRAModel=BERT, rank=22026.046091.999.698—
Same-layer LoRAModel=DistilBERT, rank=22026.0456.775.99997.5—
All-layers Full FTModel=BERT2026.0456.710099.593.8—
All-layers Full FTModel=DistilBERT2026.0450.210099.593.6—
FPVictim=Llama2-7B, Attack=Cbat2025.1150.14———80.97
FPVictim=Llama2-7B, Attack=HiddenKiller2025.1148.27———80.1
FPVictim=Llama2-7B, Attack=BadNets2025.1146.39———83.19
PUREVictim=Llama2-7B, Attack=HiddenKiller2025.1140.24———79.39
PUREVictim=Llama2-7B, Attack=Cbat2025.1139.49———80.35
FPVictim=Llama2-7B, Attack=BGMAttack2025.1139.33———82.78
FPVictim=BERT, Attack=Cbat2025.1139.23———89.93
MEFTVictim=BERT, Attack=HiddenKiller2025.1138.05———89.21
FPVictim=BERT, Attack=BGMAttack2025.1137.15———86.17
FPVictim=BERT, Attack=HiddenKiller2025.1136.45———89.17
MEFTVictim=BERT, Attack=Cbat2025.1136.12———91.06
PUREVictim=Llama2-7B, Attack=BGMAttack2025.1135.45———80.28
PUREVictim=BERT, Attack=HiddenKiller2025.1135.26———88.39
PUREVictim=Llama2-7B, Attack=BadNets2025.1133.45———82.37
FPVictim=Llama2-7B, Attack=NWS2025.1132.04———81.45
PUREVictim=BERT, Attack=BGMAttack2025.1129.46———81.27
PUREVictim=Llama2-7B, Attack=NWS2025.1128.44———83.61
PUREVictim=BERT, Attack=Cbat2025.1128.23———89.76
MEFTVictim=Llama2-7B, Attack=HiddenKiller2025.1128.17———80.78
FPVictim=BERT, Attack=BadNets2025.1126.36———92.28
MEFTVictim=Llama2-7B, Attack=BadNets2025.1126.34———85.17
Attention Safety AlignmentVictim=Llama2-7B, Attack=BadNets2025.1125.61———85.19
FPVictim=BERT, Attack=NWS2025.1125.33———87.19
MEFTVictim=Llama2-7B, Attack=BGMAttack2025.1124.03———81.3
MEFTVictim=Llama2-7B, Attack=NWS2025.1123.91———82.19
OriginalModel=BERT2026.0423.2198.298.5—
Attention Safety AlignmentVictim=Llama2-7B, Attack=Cbat2025.1123.14———83.19
Attention Safety AlignmentVictim=Llama2-7B, Attack=BGMAttack2025.1123.07———82.95
MEFTVictim=BERT, Attack=BGMAttack2025.1122.75———84.31
MEFTVictim=Llama2-7B, Attack=Cbat2025.1122.55———80.13
OriginalModel=DistilBERT2026.0422.35.698.198.2—
Attention Safety AlignmentVictim=BERT, Attack=BGMAttack2025.1121.73———85.58
Attention Safety AlignmentVictim=BERT, Attack=Cbat2025.1120.36———91.45
Attention Safety AlignmentVictim=Llama2-7B, Attack=HiddenKiller2025.1120.16———82.04
PUREVictim=BERT, Attack=NWS2025.1120.15———88.3
MEFTVictim=BERT, Attack=NWS2025.1119.67———89.31
Attention Safety AlignmentVictim=Llama2-7B, Attack=NWS2025.1118.33———83.16
PUREVictim=BERT, Attack=BadNets2025.1116.98———90.64
Attention Safety AlignmentVictim=BERT, Attack=HiddenKiller2025.1115.66———90.03
Attention Safety AlignmentVictim=BERT, Attack=NWS2025.1115.48———90.54
MEFTVictim=BERT, Attack=BadNets2025.1115.23———91.48
Attention Safety AlignmentVictim=BERT, Attack=BadNets2025.1111.23———91.09