Can I read Genixer: Visual Instruction Data Generation on EtoBox?
Genixer: Visual Instruction Data Generation by meggykharatyan793 is a document available to read on EtoBox.
What is Genixer: Visual Instruction Data Generation about?
The paper introduces Genixer, a novel data generation pipeline for Multimodal Large Language Models (MLLMs) that enables them to independently generate visual instruction tuning data without relying on GPT-4. The pipeline consists of four steps: instruction data collection, template design, empowering MLLMs, and data generation and filtering, leading to the creation of two high-quality datasets, Genixer-915K and Genixer-350K. Experimental results demonstrate that MLLMs trained with these datasets outperform
- Author
- meggykharatyan793
- Language
- EN