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

SafeMT

Benchmarks

Task NameDataset NameSOTA ResultTrend
Safety EvaluationSafeMT
Helpfulness82.1
12
Multi-turn decomposition defensesafemt
ASR2
7
Adversarial AttackSafeMT ATTACK 600
Attack Success Rate71.29
4
Showing 3 of 3 rows