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

Safety Alignment on AdvBench

-0.38Reward

SEA

-12.9016-9.6508-6.4-3.1492May 26, 2025
Updated 1mo ago

Evaluation Results

MethodLinks
2025.05
-0.38---0.1975.32
2025.05
-1.51---0.96-24.68
2025.05
-1.55---0.770
2025.05
-1.75---0.96-24.68
2025.05
-1.83---3.8573.3
2025.05
-2.11---0.96-24.68
2025.05
-2.36---0.77-
2025.05
-2.45---0.3850.65
2025.05
-3.84---6.3555.96
2025.05
-4.03---6.9286.37
2025.05
-4.29---8.8538.63
2025.05
-4.96---0.1975.32
2025.05
-5---8.6540.01
2025.05
-5.41---8.2742.65
2025.05
-5.61---5.5891.54
2025.05
-6.13---28.2744.32
2025.05
-6.32---11.7318.65
2025.05
-6.86---28.2744.32
2025.05
-7.16---43.8533.52
2025.05
-7.41---658.39
2025.05
-7.62---23.6553.42
2025.05
-7.97---22.555.68
2025.05
-8.07---43.6533.82
2025.05
-8.1---14.42-
2025.05
-8.24---24.8162.38
2025.05
-8.48---32.1236.73
2025.05
-8.76---25.9660.64
2025.05
-9.59---49.2325.36
2025.05
-9.98---4021.21
2025.05
-10.73---5024.2
2025.05
-11.95---50.77-
2025.05
-12.42---65.96-
2025.06
-33616--
2025.06
-33643--
2025.06
-9901--
2025.06
-9910--
2025.06
-9910--