benchgap
Calibration

Artificial Analysis Decision Index v1.0 → WildGuardTest (Decision Index)

WildGuardTest (Decision Index) is estimated from Artificial Analysis Decision Index v1.0 with a linear curve fitted on 19 models measured on both: y = 0.3071·x + 0.7501, R² = 0.43, cross-validated error 2.5 pp. It is used for 0 estimates.

Fitted on

ModelArtificial Analysis Decision Index v1.0WildGuardTest (Decision Index)
Celeris-1.1 Decision60.9%94.9%
Celeris-1 Decision52.3%93.3%
Clef48.6%90.7%
Clef-Flash45.8%90.9%
Clef-Omni44.3%87.2%
d1-ultra65.2%95.4%
Drex Max63.7%95.5%
Drex Max v1.163.8%95.5%
Drex v1.550.7%86.9%
Drex v2.057.9%87.9%
elephant-164.7%93.3%
GLiDE53.9%95.1%
GLiDE v1.265.0%95.4%
Jev 1.1457.1%93.1%
d150.3%90.9%
Mercury Decide52.7%88.0%
Microsoft-Decision-153.8%88.2%
Perplexity Decider Flash v1.2 4B50.2%93.3%
Perplexity Decider v1.2 27B56.2%94.4%