List your company
ResearchFormerly WizWand

Helpfulness Evaluation on MM-Vet2 (test)

54.4GPT-Eval ScoreNo Defense
10.61621.98333.3544.717Aug 26, 2025
Updated 7mo ago

Evaluation results

MethodLinks
No DefenseBackbone=Qwen2-VL2025.0854.4
PRISMBackbone=Qwen2-VL2025.0848.9
SPA-VLBackbone=Qwen2-VL2025.0846.8
PRISMBackbone=LLaVA-1.52025.0820.4
SPA-VLBackbone=LLaVA-1.52025.0820.2
SafeRLHF-VBackbone=LLaVA-1.52025.0819.3
VLGuardBackbone=Qwen2-VL2025.0817.7
No DefenseBackbone=LLaVA-1.52025.0813.1
SafeRLHF-VBackbone=Qwen2-VL2025.0812.9
VLGuardBackbone=LLaVA-1.52025.0812.3