LLAVA-RLHF FOR VISUALLY ALIGNED MODELS

Summary from the TLDR AI newsletter

Llava is a method that inserts vision capabilities into language models. It has now been instruction-tuned and run through an RLHF process. Llava improves instruction following and reasoning abilities.