Forgotten Polygons: Multimodal Large Language Models Are Shape-Blind

2 points| chbint | 1 month ago |arxiv.org