Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Jailbreak on AdvBench 50 most harmful requests
Loading...
95.83
Attack Success Rate (ASR)
MIDAS
-3.8332
22.0409
47.915
73.7891
Feb 28, 2026
Attack Success Rate (ASR)
Harm Rate (HR)
Updated 1mo ago
Evaluation Results
Method
Method
Links
Attack Success Rate (ASR)
Harm Rate (HR)
MIDAS
Target Model=QVQ-Max
2026.02
95.83
4.19
MIDAS
Target Model=Gemini-2....
2026.02
90
4.57
MIDAS
Target Model=GPT-4o
2026.02
80
3.12
MIDAS
Target Model=GPT-5-Chat
2026.02
64
3.02
FigStep
Target Model=QVQ-Max
2026.02
30.61
1.04
HIMRD
Target Model=Gemini-2....
2026.02
18.3
0.71
HIMRD
Target Model=GPT-4o
2026.02
12
0.54
HIMRD
Target Model=QVQ-Max
2026.02
10.2
0.67
FigStep
Target Model=Gemini-2....
2026.02
0
0.6
FigStep
Target Model=GPT-4o
2026.02
0
0.35
FigStep
Target Model=GPT-5-Chat
2026.02
0
0
HIMRD
Target Model=GPT-5-Chat
2026.02
0
0.08
Feedback
Search any
task
Search any
task