List your company

Speculative Advice Attack on SecLens Human Verified (test)

48.2Attack Success RateSecLens
4.228815.644427.0638.4756Jun 14, 2025

Evaluation results

MethodLinks
SecLensModel=GPT-4o2025.0648.2
SecLensModel=GPT-4-turbo2025.0645.62
MCTSModel=GPT-4o2025.0644.97
MCTSModel=GPT-4-turbo2025.0641.91
SecLensModel=Claude-3.72025.0641.72
MCTSModel=Claude-3.72025.0637.06
TuningModel=GPT-4o2025.0636.74
TuningModel=GPT-4-turbo2025.0631.84
TuningModel=Claude-3.72025.0630.11
RandomModel=GPT-4-turbo2025.0612.03
RandomModel=GPT-4o2025.0611.93
RandomModel=Claude-3.72025.065.92