List your company

Reasoning on ECQA

83.96CACCMirageBD
68.505672.517876.5380.5422Apr 8, 2026

Evaluation results

MethodLinks
MirageBDBackbone=Qwen2.5-7B, Poison Ratio=0.052026.0483.9699.91
BenignBackbone=Qwen2.5-7B, Poison Ratio=0.052026.0483.09—
BadChainBackbone=Qwen2.5-7B, Poison Ratio=0.052026.0482.950.09
SFT+RLBackbone=Qwen2.5-7B, Poison Ratio=0.052026.0482.91100
DecepChainBackbone=Qwen2.5-7B, Poison Ratio=0.052026.0481.2799.59
BadChainBackbone=Llama3.1-8B, Poison Ratio=0.052026.0479.810
MirageBDBackbone=Llama3.1-8B, Poison Ratio=0.052026.0479.499.36
DecepChainBackbone=Qwen2.5-3B, Poison Ratio=0.052026.0479.260
BenignBackbone=Qwen2.5-3B, Poison Ratio=0.052026.0478.71—
MirageBDBackbone=Qwen2.5-3B, Poison Ratio=0.052026.0477.6299.82
SFT+RLBackbone=Qwen2.5-3B, Poison Ratio=0.052026.0476.6297.04
BadChainBackbone=Qwen2.5-3B, Poison Ratio=0.052026.0476.210
BenignBackbone=Llama3.2-3B, Poison Ratio=0.052026.0474.48—
DecepChainBackbone=Llama3.2-3B, Poison Ratio=0.052026.0474.2598.27
BadChainBackbone=Llama3.2-3B, Poison Ratio=0.052026.0473.880
MirageBDBackbone=Llama3.2-3B, Poison Ratio=0.052026.0473.4399.82
BadChainBackbone=Qwen2.5-1.5B, Poison Ratio=0.052026.0473.290
BenignBackbone=Llama3.1-8B, Poison Ratio=0.052026.0473.2—
MirageBDBackbone=Qwen2.5-1.5B, Poison Ratio=0.052026.0472.6199.45
DecepChainBackbone=Qwen2.5-1.5B, Poison Ratio=0.052026.0472.4299.91
BenignBackbone=Qwen2.5-1.5B, Poison Ratio=0.052026.0471.51—
SFT+RLBackbone=Llama3.2-3B, Poison Ratio=0.052026.0471.193.98
SFT+RLBackbone=Qwen2.5-1.5B, Poison Ratio=0.052026.0470.51100
SFT+RLBackbone=Llama3.1-8B, Poison Ratio=0.052026.0469.780
DecepChainBackbone=Llama3.1-8B, Poison Ratio=0.052026.0469.10