GeoReader: Learning Latent Geometry for Spatial Reasoning in Vision-Language Models
Vision-language models (VLMs) struggle with metric and multi-view relationships that remain implicit in RGB observations. Geometry foundation models recover this structure from RGB, but existing methods either fuse the estimated features with visual tokens and learn to use them from question answering alone, or align V...
Rong Li, Shi-Jie Li, Jiangyong Huang et al.
· 0 citations