← Flash PapersVisual Instruction Tuning: Demystifying LLaVA

Share this Flash Paper

Visual Instruction Tuning: Demystifying LLaVA

https://flashpapers.ai/p/5bvqdwjqc6
Make your own
Haotian Liu, Chunyuan Li +2 more
Details ▸
ScienceQA Accuracy:92.53%· ScienceQARelative Score vs GPT-4:85.1%· LLaVA-Bench (COCO)Relative Score vs GPT-4 (Complex Reasoning):96.5%· LLaVA-Bench (COCO)