ALLaVA: Harnessing GPT4V-synthesized Data for A Lite Vision-Language Model

arXiv:2402.11684v1 Announce Type: new
Abstract: Recent advancements in Large Vision-Language Models (LVLMs) have enabled processing of multimodal inputs in language models but require significant computational resources for deployment, especially in edge devices. This study aims to bridge the performance gap between traditional-scale LVLMs and resource-friendly lite versions by adopting high-quality training data. To do this, a synthetic dataset is created by leveraging GPT-4V's ability to generate detailed captions, complex reasoning instructions and detailed answers from images. The resulted model trained …

arxiv cs.ai cs.cl data gpt4v language language model synthesized type vision