List your company

Prompt Injection Attack on Direct Scenario

2.88ASRStruQ
-0.695623.439747.57571.7103Apr 29, 2025

Evaluation results

MethodLinks
StruQBackbone=Llama3-8B-Instruct, Attack Method=GCG2025.042.88
Robustness via ReferencingBackbone=Llama3-8B-Instruct, Attack Method=GCG2025.043.85
StruQBackbone=Llama3-8B-Instruct, Attack Method=AutoDAN2025.045.77
Robustness via ReferencingBackbone=Llama3-8B-Instruct, Attack Method=AutoDAN2025.046.25
Robustness via ReferencingBackbone=Qwen2-7B-Instruct, Attack Method=GCG2025.046.25
Robustness via ReferencingBackbone=Qwen2-7B-Instruct, Attack Method=AutoDAN2025.048.65
StruQBackbone=Qwen2-7B-Instruct, Attack Method=GCG2025.0410.1
StruQBackbone=Qwen2-7B-Instruct, Attack Method=AutoDAN2025.0414.42
SpotlightBackbone=Llama3-8B-Instruct, Attack Method=GCG2025.0417.79
SandwichBackbone=Llama3-8B-Instruct, Attack Method=GCG2025.0424.04
InstructionalBackbone=Llama3-8B-Instruct, Attack Method=GCG2025.0425.48
SpotlightBackbone=Llama3-8B-Instruct, Attack Method=AutoDAN2025.0425.48
ReminderBackbone=Llama3-8B-Instruct, Attack Method=GCG2025.0427.88
SandwichBackbone=Llama3-8B-Instruct, Attack Method=AutoDAN2025.0431.25
SandwichBackbone=Qwen2-7B-Instruct, Attack Method=GCG2025.0432.21
InstructionalBackbone=Llama3-8B-Instruct, Attack Method=AutoDAN2025.0439.9
SandwichBackbone=Qwen2-7B-Instruct, Attack Method=AutoDAN2025.0440.38
SpotlightBackbone=Qwen2-7B-Instruct, Attack Method=GCG2025.0441.83
ReminderBackbone=Llama3-8B-Instruct, Attack Method=AutoDAN2025.0443.75
SpotlightBackbone=Qwen2-7B-Instruct, Attack Method=AutoDAN2025.0447.12
InstructionalBackbone=Qwen2-7B-Instruct, Attack Method=GCG2025.0457.69
NoneBackbone=Llama3-8B-Instruct, Attack Method=AutoDAN2025.0460.58
ReminderBackbone=Qwen2-7B-Instruct, Attack Method=GCG2025.0472.6
InstructionalBackbone=Qwen2-7B-Instruct, Attack Method=AutoDAN2025.0472.6
ReminderBackbone=Qwen2-7B-Instruct, Attack Method=AutoDAN2025.0485.58
NoneBackbone=Qwen2-7B-Instruct, Attack Method=GCG2025.0488.94
NoneBackbone=Qwen2-7B-Instruct, Attack Method=AutoDAN2025.0489.42
NoneBackbone=Llama3-8B-Instruct, Attack Method=GCG2025.0492.27