List your company
ResearchFormerly WizWand

Multimodal Benchmarking on MMBench

83.4ScoreLongVILA-7B (S3)
3.52824.2644565.736Nov 16, 2023Apr 15, 2024Sep 14, 2024Feb 12, 2025Jul 14, 2025Dec 12, 2025May 13, 2026
Updated 5mo ago

Evaluation results

MethodLinks
LongVILA-7B (S3)LLM=Qwen2-7B, Resolution=dynamic2024.0883.4
Meteor2024.0582.9
Brote-IM-XXL#Param LLM=11B, Setting=Zero-shot, Input=Single-image2024.0277.34
Brote-EX-XXL#Param LLM=11B, Setting=Zero-shot, Input=Single-image2024.0276.67
MMICL-XXL#Param LLM=11B, Setting=Zero-shot, Input=Single-image2024.0276.58
Cambrian-1-8BLLM=Llama 3-8B, Resolution=10242024.0875.9
INF-LLaVA*Source=Ours, larger_dataset=true2024.0774.38
Brote-IM-XL#Param LLM=3B, Setting=Zero-shot, Input=Single-image2024.0274.29
GRACEData %=20%2026.0573.8
StepmaxData %=20%2026.0573.7
LongestData %=20%2026.0573.6
Brote-EX-XL#Param LLM=3B, Setting=Zero-shot, Input=Single-image2024.0273.27
MMICL-XL#Param LLM=3B, Setting=Zero-shot, Input=Single-image2024.0273.11
CADCData %=20%2026.0573.1
Mini-Gemini-HD-8BLLM=Llama 3-8B, Resolution=15362024.0872.7
RandomData %=20%2026.0572.6
ICONSData %=20%2026.0571.5
SPHINX-MoELLM Training #P=8x7B, Input image resolution (Res.)=448, Pre-training samples (PT)=-, Instruction tuning samples (IT)=15.3M2024.0871.3
GRACEData %=5%2026.0571.1
GRACEData %=15%2026.0571.1
INF-LLaVASource=Ours2024.0770.35
InstructBLIP-XXL#Param LLM=11B, Setting=Zero-shot, Input=Single-image2024.0270.34
VILALLM=Llama 2-13B, Resolution=3362024.0870.3
LESSData %=20%2026.0570.2
HoneybeeSource=CVPR’242024.0770.1
InstructBLIP-XL#Param LLM=3B, Setting=Zero-shot, Input=Single-image2024.0269.68
Prompt HighlighterSource=CVPR’242024.0769.5
GRACEData %=10%2026.0569.5
FullData %=100%2026.0569.3
VILALLM=Llama 2-7B, Resolution=3362024.0868.9
ConvLLaVASource=arXiv’242024.0768.7
AcFormerSource=arXiv’242024.0768.4
MoE-LLaVA-2.7Bx4LLM Training #P=5B, Input image resolution (Res.)=384, Pre-training samples (PT)=558K, Instruction tuning samples (IT)=1.6M2024.0868
MoExtendLLM Training #P=3B, Input image resolution (Res.)=336, Pre-training samples (PT)=558K, Instruction tuning samples (IT)=665K2024.0867.8
LLaVA-1.5LLM Training #P=13B, Input image resolution (Res.)=336, Pre-training samples (PT)=558K, Instruction tuning samples (IT)=665K2024.0867.7
LLaVA-1.5LLM=Vicuna-1.5-13B, Resolution=3362024.0867.7
HyperLLaVALLM Training #P=7B, Input image resolution (Res.)=336, Pre-training samples (PT)=558K, Instruction tuning samples (IT)=665K2024.0865.9
TIVESource=arXiv’242024.0765.8
RoboMamba2024.0565.7
MMICL-XXL (BLIP-2)#Param LLM=11B, Setting=Zero-shot, Input=Single-image2024.0265.24
MoE-LLaVA-2.7Bx4LLM Training #P=5B, Input image resolution (Res.)=336, Pre-training samples (PT)=558K, Instruction tuning samples (IT)=1.6M2024.0865.2
GenLLaVASource=arXiv’242024.0765
LLaVA-NeXT-8BLLM=Llama 3-8B, Resolution=6722024.0864.6
LLaVA1.5Source=CVPR’242024.0764.3
LLaVA-1.5LLM Training #P=7B, Input image resolution (Res.)=336, Pre-training samples (PT)=558K, Instruction tuning samples (IT)=665K2024.0864.3
LLaVA-1.5LLM=Vicuna-1.5-7B, Resolution=3362024.0864.3
InstructBLIPLLM=V-7B, Res.=2242023.1160.9
Video-LLaVALLM=V-7B, Res.=2242023.1160.9
Qwen-VL-7B-ChatLLM Training #P=7B, Input image resolution (Res.)=448, Pre-training samples (PT)=1.4B, Instruction tuning samples (IT)=50M2024.0860.6
Qwen-VL-ChatLLM=Qwen-7B, Resolution=4482024.0860.6
OneLLMSource=CVPR’242024.0760
LLaVA-PhiSource=arXiv’242024.0759.8
LLaVA-1.5†LLM=V-7B, Res.=224, Encoder=LanguageBind-Image2023.1159.5
ShikraSource=arXiv’232024.0758.8
ShikraLLM Training #P=13B, Input image resolution (Res.)=224, Pre-training samples (PT)=600K, Instruction tuning samples (IT)=5.5M2024.0858.8
OtterHD-8BSource=arXiv’232024.0758.3
DreamLLMSource=ICLR’242024.0758.2
InstructBLIPLLM=V-13B, Res.=2242023.1158.2
VL-Mamba2024.0557
IDEFICS-80BLLM Training #P=65B, Input image resolution (Res.)=224, Pre-training samples (PT)=353M, Instruction tuning samples (IT)=1M2024.0854.5
Otter#Param LLM=7B, Setting=Zero-shot, Input=Single-image2024.0248.3
IDEFICS-9BLLM Training #P=7B, Input image resolution (Res.)=224, Pre-training samples (PT)=353M, Instruction tuning samples (IT)=1M2024.0848.2
LLaVASource=NeurIPS’232024.0738.7
GILLSource=NeurIPS’232024.0738.2
Qwen-VL-7BLLM Training #P=7B, Input image resolution (Res.)=448, Pre-training samples (PT)=1.4B, Instruction tuning samples (IT)=50M2024.0838.2
Qwen-VLLLM=Qwen-7B, Resolution=4482024.0838.2
BLIP-2LLM=V-13B, Res.=2242023.1138.1
LLaVA#Param LLM=7B, Setting=Zero-shot, Input=Single-image2024.0236.2
AnyGPTSource=arXiv’242024.0736
InstructBLIP-7BLLM Training #P=7B, Input image resolution (Res.)=224, Pre-training samples (PT)=129M, Instruction tuning samples (IT)=1.2M2024.0836
InstructBLIPLLM=Vicuna-7B, Resolution=2242024.0836
MiniGPT-4LLM=L-7B, Res.=2242023.1123
MGIESource=ICLR’242024.076.6