The Dual Mechanisms of Spatial Variable Binding in Vision-Language Models

arXiv:2603.22278v2 Announce Type: replace-cross Abstract: Many multimodal tasks, such as image captioning and visual question answering, require vision-language models (VLMs) to bind objects with their properties and spatial relations. Yet it remains unclear where and how such associations are computed within VLMs. In this work, we show that VLMs rely on two concurrent mechanisms to represent spatial variable binding. In the language model backbone, intermediate layers represent content-independent spatial relations on top of visual tokens corresponding to objects. However, this mechanism plays only a secondary role in shaping model predictions. Instead, the dominant source of spatial information originates in the vision encoder, whose representations encode the layout of objects and are directly exploited by the language model backbone. Notably, this spatial signal is distributed globally across visual tokens, extending beyond object regions into surrounding background areas. We show that enhancing these vision-derived spatial representations globally across all image tokens improves spatial variable binding performance across models of various sizes on complex natural images from the COCO datasets. Together, our results clarify how spatial variable binding is computed within VLMs and highlight the central role of vision encoders in enabling it.

Abstract:Many multimodal tasks, such as image captioning and visual question answering, require vision-language models (VLMs) to bind objects with their properties and spatial relations. Yet it remains unclear where and how such associations are computed within VLMs. In this work, we show that VLMs rely on two concurrent mechanisms to represent spatial variable binding. In the language model backbone, intermediate layers represent content-independent spatial relations on top of visual tokens corresponding to objects. However, this mechanism plays only a secondary role in shaping model predictions. Instead, the dominant source of spatial information originates in the vision encoder, whose representations encode the layout of objects and are directly exploited by the language model backbone. Notably, this spatial signal is distributed globally across visual tokens, extending beyond object regions into surrounding background areas. We show that enhancing these vision-derived spatial representations globally across all image tokens improves spatial variable binding performance across models of various sizes on complex natural images from the COCO datasets. Together, our results clarify how spatial variable binding is computed within VLMs and highlight the central role of vision encoders in enabling it.

Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)

arXiv:2603.22278 [cs.CV]

(or arXiv:2603.22278v2 [cs.CV] for this version)

https://doi.org/10.48550/arXiv.2603.22278 Focus to learn more arXiv-issued DOI via DataCite

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

analiza aiqbench

The Dual Mechanisms of Spatial Variable Binding in Vision-Language Models

Subiecte conexe