Enhancing Visual-Language Modality Alignment in Large Vision Language Models via Self-Improvement

Xiyao Wang, Jiuhai Chen, Zhaoyang Wang +8

Large vision-language models (LVLMs) have achieved impressive results in visual question-answering and reasoning tasks through vision instruction tuning on specific datasets. Howev…