Search papers, labs, and topics across Lattice.
This paper introduces GeoSR, a framework to improve spatial reasoning in VLMs by encouraging the models to actively use geometric information. GeoSR employs Geometry-Unleashing Masking, which strategically masks 2D vision tokens to force reliance on geometry, and Geometry-Guided Fusion, which adaptively amplifies geometry token contributions. Experiments on static and dynamic spatial reasoning benchmarks show that GeoSR outperforms existing methods, demonstrating the effectiveness of leveraging geometric information.
VLMs aren't using 3D geometry tokens effectively for spatial reasoning, but a simple masking and gated fusion strategy can unlock significant performance gains.
Empowered by large-scale training, vision-language models (VLMs) achieve strong image and video understanding, yet their ability to perform spatial reasoning in both static scenes and dynamic videos remains limited. Recent advances try to handle this limitation by injecting geometry tokens from pretrained 3D foundation models into VLMs. Nevertheless, we observe that naive token fusion followed by standard fine-tuning in this line of work often leaves such geometric cues underutilized for spatial reasoning, as VLMs tend to rely heavily on 2D visual cues. In this paper, we propose GeoSR, a framework designed to make geometry matter by encouraging VLMs to actively reason with geometry tokens. GeoSR introduces two key components: (1) Geometry-Unleashing Masking, which strategically masks portions of 2D vision tokens during training to weaken non-geometric shortcuts and force the model to consult geometry tokens for spatial reasoning; and (2) Geometry-Guided Fusion, a gated routing mechanism that adaptively amplifies geometry token contributions in regions where geometric evidence is critical. Together, these designs unleash the potential of geometry tokens for spatial reasoning tasks. Extensive experiments on both static and dynamic spatial reasoning benchmarks demonstrate that GeoSR consistently outperforms prior methods and establishes new state-of-the-art performance by effectively leveraging geometric information. The project page is available at https://suhzhang.github.io/GeoSR/.
