Can Large Vision-Language Models Correct Semantic Grounding Errors By Themselves? | IEEE Conference Publication | IEEE Xplore