Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

System Prompt Extraction on Instruction Hierarchy (test)

99.7Attack Success Rate (Realistic User)

OpenAI o3

88.05291.07694.197.124Dec 19, 2025
Updated 4d ago

Evaluation Results

MethodLinks
2025.12
99.798.298.2
9999.199.1
2025.12
88.59378.9
2025.12
88.582.556.1