arXiv 2606.089522026-06-08cs.AI

AlloSpatial：基础模型中空间推理的智能体框架

AlloSpatial: Agentic Harness Framework for Spatial Reasoning in Foundation Models

Institute of Artificial Intelligence, Beihang University（北京航空航天大学人工智能研究院）；Huawei Noah’s Ark Lab（华为诺亚方舟实验室）；University of Science and Technology Beijing（北京科技大学）

Shouwei Ruan, Bin Wang, Zhenyu Wu, Qihui Zhu, Yuxiang Zhang, Jingzhi Li, Yubin Wang, Xingxing Wei

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

提出AlloSpatial框架，通过World2Mind认知映射沙箱将自我中心观察转化为异中心空间先验，并利用空间推理工具实现几何语义仲裁，在VSI-Bench和MindCube上提升模型5%-18%的空间推理性能。

摘要：

多模态基础模型（MFMs）取得了显著进展，但在物理世界的空间推理中仍然脆弱。一个关键瓶颈在于它们无法将局部的自我中心观察转化为全局的异中心空间表示。为了解决这个问题，我们提出了AlloSpatial，一个用于基础模型中异中心空间认知的智能体框架。AlloSpatial引入了World2Mind，一个即插即用的认知映射沙箱，将自我中心观察转化为结构化的异中心先验，包括异中心空间树和路线图，支持查询对象拓扑、几何关系、可通过性和轨迹。为了在噪声重建和模糊视觉证据下可靠地利用这些先验，AlloSpatial引入了空间推理工具，用于工具使用判断、模态解耦线索收集和几何语义仲裁。我们进一步通过冷启动强化学习，使用工具门控轨迹级奖励，在Qwen3-VL中内化这一过程。在VSI-Bench和MindCube上的实验表明，AlloSpatial在无训练设置下将专有模型提升了5%-18%，而仅ASTs就在移除视觉输入时支持强大的空间推理。训练后的AlloSpatial智能体进一步超越了更大的通用模型和竞争性的空间基线，表明结构化的异中心表示、主动工具使用和可验证推理为具有空间能力的基础模型提供了一条有前景的路径。

英文摘要：

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.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。