做 3D 视觉或机器人空间推理的团队，终于有了一个无需微调就能显著提升 VLM 空间理解能力的框架——SpatialClaw 用代码接口解决了灵活性和中间反馈的痛点，值得在现有模型上直接试试。
SpatialClaw 提出了一种无需训练的空间推理框架，通过将代码作为动作接口，让 VLM 智能体能够逐步执行并观察中间结果，从而灵活组合感知和几何操作。现有空间智能体要么单次执行代码（无法中途调整），要么依赖结构化工具调用（灵活性不足），限制了复杂 3D/4D 推理。SpatialClaw 维护一个状态化 Python 内核，预加载输入帧和感知原语，智能体每步写一个可执行单元，基于之前输出调整策略。在 20 个空间推理基准上，SpatialClaw 平均准确率 59.9%，比最新空间智能体高 11.2 个百分点，且跨 6 个 VLM 骨干网络表现一致。
SpatialClaw: Rethinking Action Interface for Agentic Spatial Reasoning
Spatial reasoning, the ability to determine where objects are, how they relate, and how they move in 3D, remains a fundamental challenge for vision-language models (VLMs). Tool-augmented agents attempt to address this by augmenting VLMs with specialist perception modules, yet their effectiveness is bounded by the action interface through which those tools are invoked. In this work, we study how the design of this interface shapes the agent's capacity for open-ended spatial reasoning. Existing spatial agents either employ single-pass code execution, which commits to a full analysis strategy before any intermediate result is observed, or rely on a structured tool-call interface that often offers less flexibility for freely composing operations or tailoring the analysis to each task. Both designs offer limited flexibility for open-ended, complex 3D/4D spatial reasoning. We therefore propose SpatialClaw, a training-free framework for spatial reasoning that adopts code as the action interface. SpatialClaw maintains a stateful Python kernel pre-loaded with input frames and a suite of perception and geometry primitives, letting a VLM-backed agent write one executable cell per step conditioned on all prior outputs, enabling the agent to flexibly compose and manipulate perception results and adapt its analysis to both intermediate text and visual observations and the demands of each problem. Evaluated across 20 spatial reasoning benchmarks spanning a broad range of static and dynamic 3D/4D spatial reasoning tasks, SpatialClaw achieves 59.9% average accuracy, outperforming the recent spatial agent by +11.2 points, with consistent gains across six VLM backbones from two model families without any benchmark- or model-specific adaptation.