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

Sneakyprompt

Benchmarks

Task NameDataset NameSOTA ResultTrend
Text-to-Image Safety GuardingSneakyPrompt-P
Unsafe Ratio1.51
9
Text-to-Image Safety GuardingSneakyPrompt-N
Unsafe Ratio0
9
Safe Generation RateSneakyprompt
GPT-4o0.7962
8
Prompt-Image AlignmentSneakyprompt
CLIPScore0.7211
8
Harmful prompt detectionSneakyPrompt
Precision100
6
Safe Image GenerationSneakyPrompt
DSR98.75
6
Safety Filter BypassSneakyprompt
NSFW-TC Score6.63
1
Showing 7 of 7 rows