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

Jailbreak Defense on MultiJail

0ASR

SelfGrader

Updated 15d ago

Evaluation Results

Method	Links
SelfGrader 2026.04		0	0
GradientCuff 2026.04		0.63	1.9
Token Highlighter 2026.04		1.26	19
GradientCuff 2026.04		2.22	25.71
GuardReasoner (Post) 2026.04		2.22	95.23
SelfDefend (Intent) 2026.04		3.17	57.77
GuardReasoner 2026.04		3.2	80.63
WildGuard (Post) 2026.04		3.49	97.77
GuardReasoner (Pre) 2026.04		3.49	35.55
GuardReasoner 2026.04		3.5	36.19
SelfDefend 2026.04		3.8	55.56
SelfDefend (Direct) 2026.04		4.12	72.38
LLama-3-8B-Instruct (No Defense) 2026.04		4.44	-
Perplexity Filter 2026.04		4.44	100
Prompt Guard 2026.04		4.44	100
Llama Guard (Pre) 2026.04		4.44	95.23
Llama Guard (Post) 2026.04		4.44	99.68
WildGuard (Pre) 2026.04		4.44	79.68
SelfGrader 2026.04		4.44	52.38
Qwen2.5-7B-Instruct (No Defense) 2026.04		5.4	-
Perplexity Filter 2026.04		5.4	100
GradientCuff 2026.04		5.4	88.25
Prompt Guard 2026.04		5.4	100
Llama Guard 2026.04		5.4	95.24
Llama Guard 2026.04		5.4	100
SelfDefend 2026.04		5.4	72.7
WildGuard 2026.04		5.4	80
WildGuard 2026.04		5.4	99.05
SelfGrader 2026.04		11.74	36.82
SelfDefend 2026.04		13.65	59.68
SelfDefend 2026.04		16.83	70.79
Llama Guard 2026.04		24.76	97.14
Llama Guard 2026.04		25.08	95.24
Vicuna-13B-v1.5 2026.04		25.4	-
Perplexity Filter 2026.04		25.4	100
Prompt Guard 2026.04		25.4	100