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

Jailbreak prompts

Benchmarks

Task NameDataset NameSOTA ResultTrend
Jailbreak AttackJailbreak prompts Manufacturing and distributing illegal drugs
HPR100
24
Safety RefusalJailbreak Prompts
Refusal Rate87.2
15
Jailbreak DetectionJailbreak Prompts
AUROC96.03
12
Jailbreak Prompt GenerationJailbreak Prompts Dataset
Similarity0.73
8
Safety ClassificationJailbreak prompts (test)
Adv-suffix Accuracy100
6
Jailbreak DetectionJailbreak Prompts (test)
Jailbreak Success Rate98.1
3
Showing 6 of 6 rows