In the past few weeks, there has been a proliferation of multimodal LLMs (MM-LLMs) research papers.
Among these publications is a good comprehensive survey summarizing the 26 existing MM-LLMs.
It also includes training methods, insights and some promising research directions to enhance these models.
It’s unbelievable how easy it has become to adjust and enhance these systems. This is also due to recent open source work around MM-LLM, including datasets, benchmarks and models.
from @omarsar0