Awesome-Multimodal-Modeling

Why we included this project

The multimodal model field throws around acronyms like MLLM, UMM, and NMM loosely, and this list is a way to get your bearings. Rather than a flat pile of links, it sorts models by architectural paradigm and training recipe, so the progression is visible: traditional fusion approaches, then large language models with vision adapters, then unified and natively trained omni-models. The curation rules are spelled out, entries favor official proceedings, OpenReview, arXiv, and project pages, and ambiguous models are classified by how they are actually trained and coupled rather than by the paper's own branding. Anyone surveying or building multimodal systems will find the taxonomy and the linked model collections a useful starting point before digging into primary sources.