Search papers, labs, and topics across Lattice.
This paper introduces AlloSpatial, an innovative framework designed to enhance spatial reasoning in Multimodal Foundation Models (MFMs) by transforming local egocentric observations into structured allocentric representations. Utilizing World2Mind, a cognitive mapping tool, the framework generates Allocentric-Spatial Trees and route maps that facilitate querying complex spatial relationships and trajectories. Experimental results demonstrate that AlloSpatial significantly improves spatial reasoning performance by 5%-18% in a training-free context, outperforming larger models and establishing a new standard for spatial cognition in foundation models.
Structured allocentric representations can boost spatial reasoning in foundation models by up to 18%, even in the absence of visual inputs.
Multimodal Foundation Models (MFMs) have made substantial progress, yet remain fragile in spatial reasoning over the physical world. A key bottleneck lies in their inability to transform local egocentric observations into a global allocentric spatial representation. To address this, we propose AlloSpatial, an agentic framework for allocentric spatial cognition in foundation models. AlloSpatial introduces World2Mind, a plug-and-play cognitive mapping sandbox that converts egocentric observations into structured allocentric priors, including Allocentric-Spatial Trees and route maps that support querying object topology, geometric relations, passability, and trajectories. To utilize these priors reliably under noisy reconstruction and ambiguous visual evidence, AlloSpatial introduces a Spatial Reasoning Harness for tool-use judgment, modality-decoupled cue collection, and geometry-semantic arbitration. We further internalize this process in Qwen3-VL through cold-start reinforcement learning with a harness-gated trajectory-level reward. Experiments on VSI-Bench and MindCube show that AlloSpatial improves proprietary models by 5%-18% in a training-free setting, while ASTs alone support strong spatial reasoning even when visual inputs are removed. The trained AlloSpatial agents further outperform larger general-purpose models and competitive spatial baselines, suggesting that structured allocentric representations, active tool use, and verifiable reasoning offer a promising route toward spatially capable foundation models.
