LLaVA-1.5: Large Language and Vision Assistant

#17: LLMs in 2023, Llama impact grants, Visualization of transformers

LLaVA-1.5: Improved Baselines with Visual Instruction Tuning

• LLaVA is a Large Language and Vision Assistant model, similar to GPT-4V(ision)

• LLaVA-1.5 is able to achieve better baselines with simple modifications to LLaVA, namely, using CLIP-ViT-L-336px with an MLP projection and adding academic-task-oriented VQA data with simple response formatting prompts.

• Demo is available here. Access the paper and GitHub.

Image from the Paper

Video: Large Language Models in 2023

• A very good video on “Large Language Models in 2023” by Hyung Won Chung of OpenAI.

• He talks about the change in perspectives needed with the change in emergent abilities of LLMs over time, scaling of transformers based models etc.

Image from YouTube Video

And More..

• Application for Llama Impact Grants by Meta is open now to address challenges across education, environment & open innovation

• A really cool visualisation by Financial Times on “How Transformer Models Work” can be seen here.