Search papers, labs, and topics across Lattice.
This paper introduces GeoAnchor, a novel framework that enhances 3D spatial understanding by decomposing spatial information into three distinct latent components: position, direction, and geometry. By interleaving text and latent reasoning, GeoAnchor effectively captures both local evidence and global context, addressing the limitations of existing methods that rely on a single latent type. Experimental results show that GeoAnchor significantly outperforms state-of-the-art approaches on a variety of complex 3D reasoning tasks, highlighting its robustness and adaptability.
GeoAnchor achieves superior 3D spatial reasoning by integrating multiple latent representations, outperforming existing methods that struggle with complex geometric scenarios.
Although multimodal large language models (MLLMs) have achieved remarkable progress, understanding 3D spatial relationships from 2D images remains a critical challenge. Existing methods primarily rely on symbolic text tokens, which inherently lack the fidelity to represent continuous geometric information. While recent methods use latent representations to enhance reasoning, relying on a single latent type cannot adapt to the diversity of spatial tasks, leading to misalignment in complex geometric scenarios. To address these limitations, we propose GeoAnchor, an interleaved text-latent reasoning framework. GeoAnchor decomposes 3D spatial information into three complementary components: position latents for object grounding, direction latents for relational orientation, and geometry latents for scene structure. These components are recombined in a structured space to construct local evidence while capturing global context, enabling dynamic and interpretable reasoning. Furthermore, we introduce a collaborative training strategy that guides the model from local spatial perception to comprehensive 3D understanding. Extensive experiments on diverse and complex 3D reasoning tasks demonstrate that GeoAnchor outperforms the state of the art, validating its effectiveness and generalization capabilities.
