Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Safety Evaluation on HarmBench, StrongReject, WildJailbreak, XSTest Suite

68.44HarmBench Score

Initial

6.76822.77938.7954.801Jan 30, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.01
68.4466.4552.85.248.22
2026.01
67.6160.0652.84.446.22
2026.01
60.9648.8852.7518.445.25
2026.01
58.6472.8449.6045.2
2026.01
56.6438.0250.622.441.92
2026.01
54.1536.4249.31.235.27
2026.01
52.6636.151.11.235.27
2026.01
47.3457.5143.851.637.58
2026.01
46.8418.2148.858.830.68
2026.01
43.8511.8236.383.643.89
2026.01
43.6941.2139.65231.64
2026.01
41.238.9536.421.229.44
2026.01
40.3733.8737.956.429.65
2026.01
38.545.7535.7561.635.41
2026.01
38.218.31430.822.58
2026.01
37.717.3542.451.622.28
2026.01
37.387.6746.610.825.61
2026.01
35.226.7139.451.220.64
2026.01
35.054.4738.350.419.57
2026.01
33.725.7535.156.820.36
2026.01
33.063.1936.23229.8
2026.01
32.393.1932.950.817.33
2026.01
28.749.5829.2217.38
2026.01
26.914.7928.88.817.33
2026.01
24.421.9228.0537.623
2026.01
24.421.2829.256.815.44
2026.01
9.630.327.452.85.05
2026.01
9.140.327.351.24.5