On the Generalization Capacities of MLLMs for Spatial Intelligence

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Gongjie, Li, Wenhao, Qian, Quanhao, Wang, Jiuniu, Zhao, Deli, Lu, Shijian, Xu, Ran
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915840275775488
author Zhang, Gongjie
Li, Wenhao
Qian, Quanhao
Wang, Jiuniu
Zhao, Deli
Lu, Shijian
Xu, Ran
author_facet Zhang, Gongjie
Li, Wenhao
Qian, Quanhao
Wang, Jiuniu
Zhao, Deli
Lu, Shijian
Xu, Ran
contents Multimodal Large Language Models (MLLMs) that directly process RGB inputs for tasks like 3D localization and navigation have shown remarkable potential. However, we argue that these RGB-only approaches are fundamentally flawed in their ability to generalize across cameras. By ignoring camera parameters, they entangle an object's physical properties with the camera's perspective, creating an irresolvable ambiguity. We show this leads MLLMs to overfit to the training camera distribution, rather than learning true and generalizable 3D geometric principles. To address this, we propose Camera-Aware MLLM framework for spatial MLLMs. It learns generalizable spatial reasoning by: (i) injecting camera intrinsics via a dense embedding that conditions each visual token; (ii) introducing a camera-aware data augmentation strategy that synthetically varies camera parameters, forcing the model to disentangle camera properties from scene content; and (iii) distilling geometric priors from a 3D vision foundation model. Extensive experiments demonstrate that camera-aware MLLMs substantially outperform their naive counterparts, particularly in cross-camera generalization tests on spatially-grounded tasks, indicating that camera-awareness is not only beneficial but also a prerequisite for robust and generalizable spatial intelligence in MLLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2603_06704
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle On the Generalization Capacities of MLLMs for Spatial Intelligence
Zhang, Gongjie
Li, Wenhao
Qian, Quanhao
Wang, Jiuniu
Zhao, Deli
Lu, Shijian
Xu, Ran
Computer Vision and Pattern Recognition
Machine Learning
Multimodal Large Language Models (MLLMs) that directly process RGB inputs for tasks like 3D localization and navigation have shown remarkable potential. However, we argue that these RGB-only approaches are fundamentally flawed in their ability to generalize across cameras. By ignoring camera parameters, they entangle an object's physical properties with the camera's perspective, creating an irresolvable ambiguity. We show this leads MLLMs to overfit to the training camera distribution, rather than learning true and generalizable 3D geometric principles. To address this, we propose Camera-Aware MLLM framework for spatial MLLMs. It learns generalizable spatial reasoning by: (i) injecting camera intrinsics via a dense embedding that conditions each visual token; (ii) introducing a camera-aware data augmentation strategy that synthetically varies camera parameters, forcing the model to disentangle camera properties from scene content; and (iii) distilling geometric priors from a 3D vision foundation model. Extensive experiments demonstrate that camera-aware MLLMs substantially outperform their naive counterparts, particularly in cross-camera generalization tests on spatially-grounded tasks, indicating that camera-awareness is not only beneficial but also a prerequisite for robust and generalizable spatial intelligence in MLLMs.
title On the Generalization Capacities of MLLMs for Spatial Intelligence
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2603.06704