Search results for: 'ALLaVA: Harnessing GPT4V-synthesized Data for A Lite Vision-Language Model.'