← Back to Dashboard
Strategy Analysis
Deep dive into how models handle different attack vectors. We compare Direct policy violations against Adversarial attempts (jailbreaks) and measure false positives on Benign queries.
🛡️ Robustness Matrix
Comparing Policy Strictness (Direct) vs Jailbreak Resistance (Adversarial).
Ideally, models should be high in Robustness (Y) without being overly Strict (X).
Ideal (Robust & Balanced)
Vulnerable (Weak)