Haotian Liu

[NeurIPS'23 Oral] Visual Instruction Tuning (LLaVA) built towards GPT-4V level capabilities and beyond.

Open-Set Grounded Text-to-Image Generation

The first competitive instance segmentation approach that runs on small edge devices at real-time speeds.

LLaVA-Plus: Large Language and Vision Assistants that Plug and Learn to Use Skills