List your company
ResearchFormerly WizWand

Long-answer Visual Question Answering on MMMU-Pro-V

26.3AccuracyQwen2.5-VL-3B
5.29210.74616.221.654Apr 8, 2026
Updated 6mo ago

Evaluation results

MethodLinks
Qwen2.5-VL-3BModel Category=Autoregressive Vision-Language Models, Tokens/NFE=1.002026.0426.3
Intern-VL-2.5-4BModel Category=Autoregressive Vision-Language Models, Tokens/NFE=1.002026.0424.6
Fast-dVLMModel Category=Diffusion Vision-Language Models, Decoding Strategy=speculative decoding (spec.), Tokens/NFE=2.632026.0424.6
Fast-dVLMModel Category=Diffusion Vision-Language Models, Decoding Strategy=masked diffusion model decoding (MDM), Tokens/NFE=1.952026.0421.4
LLaDA-VModel Category=Diffusion Vision-Language Models, Tokens/NFE=1.002026.0418.6
DimpleModel Category=Diffusion Vision-Language Models, Tokens/NFE=1.002026.0412.4
LaViDaModel Category=Diffusion Vision-Language Models, Tokens/NFE=1.002026.0410.5
MiniCPM-V-2 (3B)Model Category=Autoregressive Vision-Language Models, Tokens/NFE=1.002026.0410.3
VILA-1.5-3BModel Category=Autoregressive Vision-Language Models, Tokens/NFE=1.002026.046.1