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

PaperGuard

Benchmarks

Task NameDataset NameSOTA ResultTrend
Prompt Injection AttackPaperGuard
Attack Success Rate (ASR)80
8
Prompt Injection DetectionPaperGuard (test)
Accuracy95
5
Adversarial Figure Manipulation DetectionPaperGuard Visual Attacks
Accuracy93.5
3
Showing 3 of 3 rows