Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Safety Alignment Evaluation on VLSBench
Loading...
100
Safety Score
GRPO
61.4368
71.4484
81.46
91.4716
Jun 2, 2026
Safety Score
Updated 1mo ago
Evaluation Results
Method
Method
Links
Safety Score
GRPO
Backbone=Qwen3-VL-4B,...
2026.06
100
COPSD (Hybrid)
Backbone=Qwen3-VL-4B,...
2026.06
99.51
Safe-RLHF-V
Backbone=Qwen3-VL-4B,...
2026.06
99.33
GRPO
Backbone=Qwen2.5-VL-7B...
2026.06
98.93
COPSD (Standard)
Backbone=Qwen3-VL-4B,...
2026.06
98.75
Safe-RLHF-V
Backbone=Qwen2.5-VL-7B...
2026.06
98.53
COPSD (Standard)
Backbone=Qwen2.5-VL-7B...
2026.06
97.72
COPSD (Hybrid)
Backbone=Qwen2.5-VL-7B...
2026.06
94.82
OPD
Backbone=Qwen2.5-VL-7B...
2026.06
90.36
SFT
Backbone=Qwen2.5-VL-7B...
2026.06
90.14
SFT
Backbone=Qwen3-VL-4B,...
2026.06
79.92
OPD
Backbone=Qwen3-VL-4B,...
2026.06
79.61
Qwen3-VL-4B
Backbone=Qwen3-VL-4B,...
2026.06
68.63
Qwen2.5-VL-7B
Backbone=Qwen2.5-VL-7B...
2026.06
62.92
Feedback
Search any
task
Search any
task