📊 New #InfinityMM dataset: 43.4M entries for training multimodal #AI models. Chinese researchers develop #AquilaVL2B model (2B parameters) achieving top scores in #MMStar (54.9%) & #MathVista (59%). #opensource model trained on #Nvidia A100 & Chinese chips. #VisionAI

arxiv.org/abs/2410.18558