LLaDA-V: Large Language Diffusion Models with Visual Instruction Tuning

Zebin You1, Shen Nie1, Xiaolu Zhang2, Jun Hu2, Jun Zhou2, Zhiwu Lu1, Ji-Rong Wen1, Chongxuan Li1
1Renmin University of China, 2Ant Group
Paper Code

TL;DR: We introduce LLaDA-V, a competitive diffusion-based vision-language model, outperforming other diffusion MLLMs.

LLaDA_vs_LLaMA
LLaDA_vs_LLaMA_chat