几何约束智能体用于空间推理
Geometrically-Constrained Agent for Spatial Reasoning
把 VLM 从「直接输出几何答案」改为「先形式化出参考系与目标约束、再在约束内调用几何工具」，这一免训练的约束式智能体范式可直接迁移到视觉空间问答、具身导航与 3D 场景理解中的几何推理管线。
原文依据 the VLM translates the user's ambiguous query into the formal, verifiable task constraint, which defines the reference frame and objective
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论视觉语言模型（VLM）在空间推理中存在根本性的语义—几何鸿沟：它们擅长定性的语义推断，但其推理运行在有损的语义空间中，与高保真几何并不对齐。现有范式无法弥合这一鸿沟。基于训练的方法受困于「神谕悖论」，从不完美的神谕中学习到有缺陷的空间逻辑。工具集成方法约束了最终计算，却关键性地让 VLM 的规划过程不受约束，从而产生几何上有缺陷的计划。在本工作中，我们提出几何约束智能体（GCA），一种免训练的智能体范式，通过引入形式化任务约束来弥合该鸿沟。具体而言，我们将 VLM 的角色策略性地解耦为两个阶段。首先，VLM 作为语义分析器，将用户模糊的查询转译为形式化、可验证的任务约束，该约束定义了参考系与目标。其次，VLM 作为任务求解器，严格在约束所定义的确定性边界内生成并执行工具调用。这一几何约束策略成功弥合了语义—几何鸿沟，为复杂空间任务给出鲁棒且可验证的推理路径。大量实验表明，GCA 在多个空间推理基准上取得 SOTA 性能，超越现有基于训练与工具集成的方法 27%。
英文原文摘要 1467 字符
Vision Language Models (VLMs) exhibit a fundamental semantic-to-geometric gap in spatial reasoning: they excel at qualitative semantic inference but their reasoning operates within a lossy semantic space, misaligned with high-fidelity geometry. Current paradigms fail to bridge this gap. Training-based methods suffer from an "oracle paradox," learning flawed spatial logic from imperfect oracles. Tool-integrated methods constrain the final computation but critically leave the VLM's planning process unconstrained, resulting in geometrically flawed plans. In this work, we propose Geometrically-Constrained Agent (GCA), a training-free agentic paradigm that resolves this gap by introducing a formal task constraint. Specifically, we strategically decouples the VLM's role into two stages. First, acting as a semantic analyst, the VLM translates the user's ambiguous query into the formal, verifiable task constraint, which defines the reference frame and objective. Second, acting as a task solver, the VLM generates and executes tool calls strictly within the deterministic bounds defined by the constraint. This geometrically-constrained strategy successfully resolve the semantic-to-geometric gap, yielding a robust and verifiable reasoning pathway for complex spatial tasks. Comprehensive experiments demonstrate that GCA achieves SOTA performance on multiple spatial reasoning benchmarks, surpassing existing training-based and tool-integrated methods by 27%.