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 VLM latents with geometry features. Capturing geometry, however, does not ensure its use: latents aligned with geometry features can be zeroed without lowering accuracy. We introduce georeader, a framework that learns geometric representations for the VLM that reads them. First, we train lightweight Geometric Readers to produce continuous input tokens, supervising them with geometry descriptions decoded by a frozen VLM, so that the Readers express geometry in a form the target model can already read. We then freeze the Readers to preserve this interface while fine-tuning the VLM on spatial QA, with geometry injection conditioned on the question. Geometry descriptions serve only as training targets: at inference, the VLM answers directly from the continuous geometric tokens without generating an intermediate geometry chain of thought. With Qwen3-VL-4B and geometry predicted from RGB images, georeader reaches $94.3$ on MindCube-Tiny and, with a multi-source QA mixture, an average of $70.6$ across six additional spatial benchmarks. These results suggest that the most useful geometric representation is the one the receiving model can use, not necessarily the one that reconstructs geometry most accurately.