List your company
ResearchFormerly WizWand

Multimodal Evaluation on MMBench-CN (test)

61.4AccuracyLLaVA-1.5 + FG-CLIP 2
-1.8611214.5624430.98647.40956Apr 25, 2024Jul 23, 2024Oct 20, 2024Jan 18, 2025Apr 17, 2025Jul 15, 2025Oct 13, 2025
Updated 5mo ago

Evaluation results

MethodLinks
LLaVA-1.5 + FG-CLIP 2Vision Encoder=FG-CLIP 2, LMM Base Architecture=LLaVA-1.52025.1061.4
LLaVA-1.5 + Meta CLIP 2Vision Encoder=Meta CLIP 2, LMM Base Architecture=LLaVA-1.52025.1060.3
LLaVA-1.5 + CLIPVision Encoder=CLIP, LMM Base Architecture=LLaVA-1.52025.1058.4
LLaVA-1.5 + SigLIP 2Vision Encoder=SigLIP 2, LMM Base Architecture=LLaVA-1.52025.1057.2
InternVL 1.5#param=26B, open-source=false2024.040.82
InternVL 1.2#param=40B, open-source=false2024.040.812
Step-1V#param=100B, open-source=false2024.040.799
LLaVA-NeXT#param=35B, open-source=false2024.040.79
HPT Proopen-source=false2024.040.767
Qwen-VL-Maxopen-source=false2024.040.757
GPT-4Vopen-source=false2024.040.744
Gemini Pro 1.0open-source=false2024.040.743
Qwen-VL-Plusopen-source=false2024.040.707
Claude-3 Sonnetopen-source=false2024.040.642
Claude-3 Opusopen-source=false2024.040.592
Claude-3 Haikuopen-source=false2024.040.572