List your company
ResearchFormerly WizWand

Large Multimodal Model Evaluation on MM-Vet

54.5Average ScoreCogVLM
8.01220.08132.1544.219Sep 14, 2023Feb 18, 2024Jul 24, 2024Dec 28, 2024Jun 3, 2025Nov 7, 2025Apr 14, 2026
Updated 6mo ago

Evaluation results

MethodLinks
CogVLMNumber of Parameters=17B, Setting=Zero-shot2024.0254.5
CogVLMZero-shot=true, Number of Parameters=17B2024.0354.5
Emu2-Chataverage of five scoring runs=true2023.1248.5
Imp-4BLLM=Phi-3-3.8B, Visual Encoder=SigLIP-SO@384, #PT=558K, #FT=1M2024.0544.6
MoAIZero-shot=true, Number of Parameters=7B2024.0343.7
Imp-3BLLM=Phi-2-2.7B, Visual Encoder=SigLIP-SO@384, #PT=558K, #FT=1M2024.0543.3
CoLLaVONumber of Parameters=7B, Setting=Zero-shot2024.0240.3
LLaVA-1.5-13B-HDLLM=Vicuna-13B, Image Size=448x448, Pre-train Sample Size=558K, Fine-tune Sample Size=665K2023.1039.4
LLaVA-1.5-13BLLM=Vicuna-1.5-13B, Visual Encoder=CLIP-L@336, #PT=558K, #FT=665K2024.0538.3
ShareGPT4VNumber of Parameters=7B, Setting=Zero-shot2024.0237.6
ShareGPT4VZero-shot=true, Number of Parameters=7B2024.0337.6
LLaVA-XTunerNumber of Parameters=20B, Setting=Zero-shot2024.0237.2
LLaVA-XTunerZero-shot=true, Number of Parameters=20B2024.0337.2
Emu-Iaverage of five scoring runs=true2023.1236.3
mPLUG-Owl2Number of Parameters=7B, Setting=Zero-shot2024.0236.2
mPLUG-Owl2-8BLLM=LLAMA2-7B, Visual Encoder=CLIP-L@448, #PT=400M, #FT=1.2M2024.0536.2
mPLUG-Owl2Zero-shot=true, Number of Parameters=7B2024.0336.2
LLaVA-1.5-13BLLM=Vicuna-13B, Image Size=336x336, Pre-train Sample Size=558K, Fine-tune Sample Size=665K2023.1036.1
LLaVA-1.5-13Baverage of five scoring runs=true2023.1235.4
LLaVA1.5Number of Parameters=13B, Setting=Zero-shot2024.0235.4
LLaVA1.5Zero-shot=true, Number of Parameters=13B2024.0335.4
LLaVA-1.5-13BImg/s=2.22, Speedup=-2025.1035.4
Intern-XCNumber of Parameters=7B, Setting=Zero-shot2024.0235.2
Intern-XCZero-shot=true, Number of Parameters=7B2024.0335.2
Merlin2023.1134.9
APTBase Model=LLaVA-1.5-13B, Img/s=2.72, Speedup=+23%2025.1034.7
ResizingBase Model=LLaVA-1.5-13B, Img/s=2.72, Speedup=+23%2025.1033.9
Imp-2BLLM=Qwen-1.5-1.8B, Visual Encoder=SigLIP-SO@384, #PT=558K, #FT=1M2024.0533.5
CLASPToken retention budget (R)=192, Base Model=LLaVA-v1.5-7B2026.0433.3
RandomBase Model=LLaVA-1.5-13B, Img/s=2.79, Speedup=+26%2025.1032
LLaVA-1.5-7BLLM=Vicuna-7B, Image Size=336x336, Pre-train Sample Size=558K, Fine-tune Sample Size=665K2023.1031.1
Mini-Gemini-2BLLM=Gemma-2B, Visual Encoder=CLIP-L@336, #PT=1.2M, #FT=1.5M2024.0531.1
APTBase Model=LLaVA-1.5-7B, Img/s=4.51, Speedup=+22%2025.1030.8
LLaVA-1.5-7BImg/s=3.70, Speedup=-2025.1030.7
LLaVA1.5Number of Parameters=7B, Setting=Zero-shot2024.0230.5
LLaVA-1.52023.1130.5
LLaVA1.5Zero-shot=true, Number of Parameters=7B2024.0330.5
NuwaToken retention budget (R)=192, Base Model=LLaVA-v1.5-7B2026.0430.5
RandomBase Model=LLaVA-1.5-7B, Img/s=4.58, Speedup=+24%2025.1030.5
LLaVA-1.5-7BLLM=Vicuna-1.5-7B, Visual Encoder=CLIP-L@336, #PT=558K, #FT=665K2024.0530.2
ResizingBase Model=LLaVA-1.5-7B, Img/s=4.51, Speedup=+22%2025.1030.2
BTPToken retention budget (R)=192, Base Model=LLaVA-v1.5-7B2026.0429.1
LLaVA-Phi-3BLLM=Phi-2-2.7B, Visual Encoder=CLIP-L@336, #PT=558K, #FT=665K2024.0528.9
LLaVANumber of Parameters=7B, Setting=Zero-shot2024.0226.7
LLaVAZero-shot=true, Number of Parameters=7B2024.0326.7
InstructBLIP-8BLLM=Vicuna-7B, Image Size=224x224, Pre-train Sample Size=129M, Fine-tune Sample Size=1.2M2023.1026.2
InstructBLIPNumber of Parameters=7B, Setting=Zero-shot2024.0226.2
InstructBLIP2023.1126.2
InstructBLIPZero-shot=true, Number of Parameters=7B2024.0326.2
InstructBLIP-13Baverage of five scoring runs=true2023.1225.6
InstructBLIP-14BLLM=Vicuna-13B, Image Size=224x224, Pre-train Sample Size=129M, Fine-tune Sample Size=1.2M2023.1025.6
InstructBLIPNumber of Parameters=13B, Setting=Zero-shot2024.0225.6
InstructBLIPZero-shot=true, Number of Parameters=13B2024.0325.6
LLaVA-7BLLM=Vicuna-7B2023.1025.5
OtterNumber of Parameters=7B, Setting=Zero-shot2024.0224.6
OtterZero-shot=true, Number of Parameters=7B2024.0324.6
MIXTraining Dataset=MIX2023.0923.9
LLaVATraining Dataset=LLaVA (2023b)2023.0923.4
BLIP-2-13Baverage of five scoring runs=true2023.1222.4
BLIP2-14BLLM=Vicuna-13B, Image Size=224x224, Pre-train Sample Size=129M2023.1022.4
BLIP2Number of Parameters=13B, Setting=Zero-shot2024.0222.4
BLIP-22023.1122.4
BLIP2Zero-shot=true, Number of Parameters=13B2024.0322.4
MiniGPT-4Number of Parameters=7B, Setting=Zero-shot2024.0222.1
MiniGPT-4Zero-shot=true, Number of Parameters=7B2024.0322.1
ShikraTraining Dataset=Shikra (2023a)2023.0919.9
TEXTBINDTraining Dataset=TEXTBIND2023.0919.4
MultiInstructTraining Dataset=MultiInstruct (2023b)2023.0917.2
MiniGPT-4Training Dataset=MiniGPT-4 (2023)2023.099.8