Building Next-Gen Multimodal Foundation Models for General-Purpose Assistants

LLaVA is an open-source project, collaborating with research community to advance the state-of-the-art in AI. LLaVA represents the first end-to-end trained large multimodal model (LMM) that achieves impressive chat capabilities mimicking spirits of the multimodal GPT-4. The LLaVA family continues growing to support more modalities, capabilities, applications and beyond.

Publications

Publications

Publications by Year

Publications by Research Area

Publications by Type