Our new X account is live! Follow @wizwand_team for updates
Home
/
Benchmarks
Jailbreak Detection on GoalFrameBench (seed prompts)
Loading...
97
Accuracy
FrameShield-Last
-0.76
24.62
50
75.38
Feb 23, 2026
Accuracy
F1 Score
Updated 4d ago
Evaluation Results
Method
Method
Links
Accuracy
F1 Score
FrameShield-Last
LLM Backbone=Vicuna-7B
2026.02
97
86
FrameShield-Last
LLM Backbone=Llama3-8B
2026.02
96
86
FrameShield-Last
LLM Backbone=Mistral-7B
2026.02
92
64
JBShield
LLM Backbone=Vicuna-7B
2026.02
91
91
JBShield
LLM Backbone=Mistral-7B
2026.02
88
88
JBShield
LLM Backbone=Qwen3-4B
2026.02
87
88
FrameShield-Last
LLM Backbone=Qwen3-4B
2026.02
84
68
JBShield
LLM Backbone=Llama3-8B
2026.02
77
86
LlamaGuard
LLM Backbone=Vicuna-7B
2026.02
75
85
LlamaGuard
LLM Backbone=Mistral-7B
2026.02
74
85
LlamaGuard
LLM Backbone=Llama3-8B
2026.02
60
75
SelfEx
LLM Backbone=Mistral-7B
2026.02
46
63
GradSafe
LLM Backbone=Llama3-8B
2026.02
37
54
SelfEx
LLM Backbone=Llama3-8B
2026.02
16
26
GradSafe
LLM Backbone=Mistral-7B
2026.02
5
10
GradSafe
LLM Backbone=Vicuna-7B
2026.02
3
6
Feedback
Search any
task
Search any
task