Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Safety Evaluation on APGD-Toxic
Loading...
3.28
ASR (Unconstrained)
ASTRA
0.5668
18.8809
37.195
55.5091
Mar 28, 2026
ASR (Unconstrained)
ASR (kappa=16/255)
ASR (kappa=32/255)
ASR (kappa=64/255)
Updated 19d ago
Evaluation Results
Method
Method
Links
ASR (Unconstrained)
ASR (kappa=16/255)
ASR (kappa=32/255)
ASR (kappa=64/255)
ASTRA
Backbone=LLaVA-OneVision
2026.03
3.28
12.4
9.75
6.62
CARE
Backbone=Qwen2.5-VL
2026.03
4.43
9.37
9.62
8.18
CARE
Backbone=LLaVA-OneVision
2026.03
4.77
9.11
8.7
7.15
ASTRA
Backbone=Qwen2.5-VL
2026.03
7.13
16.08
13.73
10.12
Original model
Backbone=Qwen2.5-VL
2026.03
67.71
59.38
61.03
70.62
Original model
Backbone=LLaVA-OneVision
2026.03
71.11
56.16
60.6
62.9
Feedback
Search any
task
Search any
task