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

Safety Classification on AdvBench

99.9F1 Score

Gradient-Controlled Decoding (GCD)

68.80476.87784.9593.023Jan 27, 2025Apr 9, 2025Jun 20, 2025Sep 1, 2025Nov 12, 2025Jan 23, 2026Apr 6, 2026
Updated 11d ago

Evaluation Results

MethodLinks
99.910099.800.19
99.910099.800.19
2026.04
99.7510099.500.5
2026.04
99.7510099.500.5
2026.04
99.6210099.2500.755
2026.04
99.6110099.2300.77
2026.04
99.2410098.501.5
2025.01
84----
2025.01
74----
2025.01
70----