List your company

Safety Evaluation on MMSafetyBench (Refusal Rate)

99.9Refusal RateSASA
18.15639.37860.681.822Jun 30, 2026

Evaluation results

MethodLinks
SASAModel=GEMMA32026.0699.9
SASAModel=QWEN3-VL2026.0699.6
SASAModel=QWEN3.52026.0699.3
SASAModel=INTERNVL 3.52026.0699.1
SASAModel=MOLMO22026.0698.4
MARSModel=QWEN3-VL2026.0685.3
MARSModel=MOLMO22026.0681.3
AdaSteerModel=QWEN3.52026.0680
MARSModel=QWEN3.52026.0676.9
AdaSteerModel=QWEN3-VL2026.0659.9
MARSModel=INTERNVL 3.52026.0655.9
ECSOModel=MOLMO22026.0646.5
Zero-shotModel=QWEN3-VL2026.0646.3
ECSOModel=QWEN3.52026.0645.5
Zero-shotModel=QWEN3.52026.0645.2
AdaSteerModel=MOLMO22026.0645.2
ECSOModel=QWEN3-VL2026.0644.8
Zero-shotModel=MOLMO22026.0639.8
MARSModel=GEMMA32026.0632.5
AdaSteerModel=INTERNVL 3.52026.0632.3
Zero-shotModel=INTERNVL 3.52026.0626.3
ECSOModel=INTERNVL 3.52026.0626.3
Zero-shotModel=GEMMA32026.0623.7
ECSOModel=GEMMA32026.0623.7
AdaSteerModel=GEMMA32026.0621.3