List your company
ResearchFormerly WizWand

Multimodal Perception Evaluation on MME-P

1,687Perception ScoreBAGEL
1,032.841,202.671,372.51,542.33Mar 1, 2026Mar 15, 2026Mar 29, 2026Apr 13, 2026Apr 27, 2026May 11, 2026May 26, 2026
Updated 5mo ago

Evaluation results

MethodLinks
BAGEL#Params=7B+7B2026.031,687
InternVL-U#Params=2B+1.7B2026.031,607.5
Qwen2.5-VL#Params=3B2026.031,574.9
MetaQuery-L#Params=3B+1B2026.031,574.3
InternVL3.5#Params=2B2026.031,552.1
LLaVA-1.5V#Params=7B2026.031,510.7
Ovis-U1#Params=2.4B+1.2B2026.031,508
OFABackbone=LLaVA-v1.5-7B, Fine-tuning Dataset=Vision-Flan-186K, Sampling Ratio=15%2026.051,499.3
XMASSampling Ratio=15% (98K samples), Base Model=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.031,485.7
Full-Data(665K)Sampling Ratio=100% (665K samples), Base Model=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.031,476.9
TUNA#Params=1.5B2026.031,461.5
PreSelSampling Ratio=15% (98K samples), Base Model=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.031,457.7
VisNecSampling Ratio=15% (98K samples), Base Model=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.031,457.2
Show-o2#Params=1.5B2026.031,450.9
Janus-Pro#Params=1.5B2026.031,444
ICONSSampling Ratio=15% (98K samples), Base Model=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.031,435.6
CoIDOSampling Ratio=15% (98K samples), Base Model=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.031,419.5
COINCIDESampling Ratio=15% (98K samples), Base Model=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.031,414.9
EL2NSampling Ratio=15% (98K samples), Base Model=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.031,405.2
RandomSampling Ratio=15% (98K samples), Base Model=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.031,397.5
TypiClustSampling Ratio=15% (98K samples), Base Model=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.031,396.2
CLIP-ScoreSampling Ratio=15% (98K samples), Base Model=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.031,380.3
Self-FilterSampling Ratio=15% (98K samples), Base Model=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.031,356.5
JanusFlow#Params=1.3B2026.031,333.1
IFDSampling Ratio=15% (98K samples), Base Model=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.031,307.2
CoIDOBackbone=LLaVA-v1.5-7B, Fine-tuning Dataset=Vision-Flan-186K, Sampling Ratio=15%2026.051,298.8
Self-FilterBackbone=LLaVA-v1.5-7B, Fine-tuning Dataset=Vision-Flan-186K, Sampling Ratio=15%2026.051,262.2
EL2NBackbone=LLaVA-v1.5-7B, Fine-tuning Dataset=Vision-Flan-186K, Sampling Ratio=15%2026.051,253.8
ICONSBackbone=LLaVA-v1.5-7B, Fine-tuning Dataset=Vision-Flan-186K, Sampling Ratio=15%2026.051,252.5
RandomBackbone=LLaVA-v1.5-7B, Fine-tuning Dataset=Vision-Flan-186K, Sampling Ratio=15%2026.051,238.6
Full-DataBackbone=LLaVA-v1.5-7B, Fine-tuning Dataset=Vision-Flan-186K, Sampling Ratio=100%2026.051,238.1
PreSelBackbone=LLaVA-v1.5-7B, Fine-tuning Dataset=Vision-Flan-186K, Sampling Ratio=15%2026.051,218.8
IFDBackbone=LLaVA-v1.5-7B, Fine-tuning Dataset=Vision-Flan-186K, Sampling Ratio=15%2026.051,210.9
TypiClustBackbone=LLaVA-v1.5-7B, Fine-tuning Dataset=Vision-Flan-186K, Sampling Ratio=15%2026.051,194.1
COINCIDEBackbone=LLaVA-v1.5-7B, Fine-tuning Dataset=Vision-Flan-186K, Sampling Ratio=15%2026.051,184.4
XMASBackbone=LLaVA-v1.5-7B, Fine-tuning Dataset=Vision-Flan-186K, Sampling Ratio=15%2026.051,151.2
CLIP-ScoreBackbone=LLaVA-v1.5-7B, Fine-tuning Dataset=Vision-Flan-186K, Sampling Ratio=15%2026.051,058