VideoMM

Researchers introduced VideoMM, an adaptive macro-micro inference framework for video multimodal large language models that decouples region selection from detailed reasoning, according to an arXiv pa…