List your company

Prompt Injection Defense on Mind2Web

84Benign UtilityVanilla
73.676.37981.7Nov 30, 2025

Evaluation results

MethodLinks
VanillaModel=Qwen3-32B2025.1184—61.246.259.848.854.363.50100
SEP DefenseModel=Qwen3-32B2025.1184—63.144.761.246.555.858.931.771.5
PromptArmorModel=Qwen3-32B2025.1184070.425.168.929.664.128.735.467.2
IntentGuardModel=Qwen3-32B2025.1181075.65.274.94.773.35.669.58.5
IntentGuardModel=gpt-oss-20B2025.1174.5069.26.168.15.866.76.464.910.9
VanillaModel=gpt-oss-20B2025.1174—53.448.352.849.748.962.131.272.6
SEP DefenseModel=gpt-oss-20B2025.1174—55.846.254.147.950.357.432.868.9
PromptArmorModel=gpt-oss-20B2025.1174061.326.659.731.156.229.633.566.1