List your company

Adversarial Attack on AdvBench Hijacked (test)

0CHRFPFT
-4235077Apr 10, 2026

Evaluation results

MethodLinks
FPFTModel=DeepSeek-7B2026.04020
OriginalModel=Llama3.2-3B2026.0411
AutoRANModel=Qwen2.5-7B2026.04819
AutoRANModel=DeepSeek-7B2026.041014
AutoRANModel=Llama3.2-3B2026.042022
CoT HijackingModel=Llama3.2-3B2026.042611
OriginalModel=Qwen2.5-7B2026.04306
OriginalModel=DeepSeek-7B2026.043219
CoT HijackingModel=DeepSeek-7B2026.044233
Chain-of-LureModel=Llama3.2-3B2026.044636
CoT HijackingModel=Qwen2.5-7B2026.044732
Chain-of-LureModel=DeepSeek-7B2026.045247
Chain-of-LureModel=Qwen2.5-7B2026.045345
H-CoTModel=Qwen2.5-7B2026.045715
H-CoTModel=Llama3.2-3B2026.04575
H-CoTModel=DeepSeek-7B2026.045933
FPFTModel=Llama3.2-3B2026.049082
LRFTModel=Llama3.2-3B2026.049399
LRFTModel=DeepSeek-7B2026.049499
LRFTModel=Qwen2.5-7B2026.049499
TSBH (Pass@5)Model=DeepSeek-7B2026.049899
FPFTModel=Qwen2.5-7B2026.049998
TSBH (Pass@1)Model=Qwen2.5-7B2026.049999
TSBH (Pass@5)Model=Qwen2.5-7B2026.049999
TSBH (Pass@1)Model=DeepSeek-7B2026.0410099
TSBH (Pass@1)Model=Llama3.2-3B2026.0410099
TSBH (Pass@5)Model=Llama3.2-3B2026.04100100