List your company

Safety Evaluation on HarmBench, StrongReject, WildJailbreak, XSTest Suite

68.44HarmBench ScoreInitial
6.76822.77938.7954.801Jan 30, 2026

Evaluation results

MethodLinks
InitialSize=0.6B2026.0168.4466.4552.85.248.22
SafePathSize=0.6B2026.0167.6160.0652.84.446.22
SafeKeySize=0.6B2026.0160.9648.8852.7518.445.25
SafeChainSize=0.6B2026.0158.6472.8449.6045.2
STAR-1Size=0.6B2026.0156.6438.0250.622.441.92
SafePathSize=1.7B2026.0154.1536.4249.31.235.27
InitialSize=1.7B2026.0152.6636.151.11.235.27
SafeChainSize=1.7B2026.0147.3457.5143.851.637.58
SafeKeySize=1.7B2026.0146.8418.2148.858.830.68
DirectRefusalSize=0.6B2026.0143.8511.8236.383.643.89
SafeChainSize=4B2026.0143.6941.2139.65231.64
SafeChainSize=8B2026.0141.238.9536.421.229.44
THINKSAFESize=0.6B2026.0140.3733.8737.956.429.65
DirectRefusalSize=1.7B2026.0138.545.7535.7561.635.41
InitialSize=4B2026.0138.218.31430.822.58
SafePathSize=4B2026.0137.717.3542.451.622.28
STAR-1Size=1.7B2026.0137.387.6746.610.825.61
SafePathSize=8B2026.0135.226.7139.451.220.64
InitialSize=8B2026.0135.054.4738.350.419.57
STAR-1Size=4B2026.0133.725.7535.156.820.36
DirectRefusalSize=4B2026.0133.063.1936.23229.8
SafeKeySize=4B2026.0132.393.1932.950.817.33
THINKSAFESize=1.7B2026.0128.749.5829.2217.38
SafeKeySize=8B2026.0126.914.7928.88.817.33
DirectRefusalSize=8B2026.0124.421.9228.0537.623
STAR-1Size=8B2026.0124.421.2829.256.815.44
THINKSAFESize=4B2026.019.630.327.452.85.05
THINKSAFESize=8B2026.019.140.327.351.24.5