List your company
ResearchFormerly WizWand

Visual Reasoning on MMMU-Pro (Discriminability)

19.5DiscriminabilitySimulatability
-0.3644.7939.9515.107Sep 30, 2025
Updated 6mo ago

Evaluation results

MethodLinks
SimulatabilityModel=Qwen2.52025.0919.5
Visual FidelityModel=LLaVA2025.0916.9
ContrastivenessModel=Qwen2.52025.0915.4
Prod(VF, Contr.)Model=Qwen2.52025.0914.6
Prod(VF, Contr.)Model=GPT-4o2025.0914.3
Min(VF, Contr.)Model=Qwen2.52025.0912.7
Min(VF, Contr.)Model=GPT-4o2025.0912.4
Avg(VF, Contr.)Model=LLaVA2025.0910.8
ContrastivenessModel=GPT-4o2025.0910.1
Avg(VF, Contr.)Model=Qwen2.52025.099.6
Avg(VF, Contr.)Model=GPT-4o2025.099.6
Visual FidelityModel=GPT-4o2025.099
Prod(VF, Contr.)Model=LLaVA2025.099
Min(VF, Contr.)Model=LLaVA2025.098.2
SimulatabilityModel=GPT-4o2025.097.9
SimulatabilityModel=LLaVA2025.095.9
ContrastivenessModel=LLaVA2025.094.7
Visual FidelityModel=Qwen2.52025.093.8
PlausibilityModel=Qwen2.52025.091.2
InformativenessModel=GPT-4o2025.091.1
InformativenessModel=LLaVA2025.090.8
InformativenessModel=Qwen2.52025.090.6
PlausibilityModel=LLaVA2025.090.6
PlausibilityModel=GPT-4o2025.090.4