Unveiling the Visual Counting Bottleneck in Vision-Language Models 文章

ArXiv CS.CV2026-05-29NEWSen作者: Xingzhou Pang, Yifan Hou, Junling Wang, Mrinmaya Sachan
查看原文 →
关系图谱

Unveiling the Visual Counting Bottleneck in Vision-Language Models · 相关技术

相关技术

visual backboneslinear probessymbolic mappingmagnitude awarenessvisual individuationvisual countingLarge vision-language model