🛡️ GF-Score: Fairness-Aware Robustness Auditing Dashboard

Inspect class-conditional adversarial robustness of certified models with four fairness metrics grounded in welfare economics. Based on GREAT Score (NeurIPS 2024), extended with per-class decomposition, disparity analysis, and attack-free self-calibration.

Metric	Meaning
RDI	Range of per-class robustness (Max Group Disparity)
NRGC	Normalized Gini Coefficient — overall inequality
WCR	Worst-case class robustness (Rawlsian maximin)
FP-GREAT	Fairness-penalized aggregate score: Ω̄ − λ·RDI