SpatialScore：迈向空间智能的全面评测
SpatialScore: Towards Comprehensive Evaluation for Spatial Intelligence
该基准覆盖 30 类空间任务与多种视觉数据形态，可直接用于诊断视觉语言模型在空间关系、视角变换上的短板；其 331K 监督微调语料与 12 个空间感知工具构成的多智能体框架，为在视觉模型中注入空间推理能力提供了可复用的数据与免训练工具编排思路。
原文依据 we propose SpatialScore, the most comprehensive and diverse multimodal spatial intelligence benchmark to date
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论现有多模态大语言模型（MLLM）在空间智能方面的评测通常零散且范围有限。本工作对现代 MLLM 的空间理解能力进行了整体评估，并提出数据驱动与智能体两类互补方案。具体而言，我们做出以下贡献：(i) 提出 SpatialScore，这是迄今最全面、最多样的多模态空间智能基准，涵盖多种视觉数据类型、输入模态与问答形式，包含跨 30 个不同任务的约 5K 条人工核验样本；(ii) 构建 SpatialCorpus，这是一个大规模训练资源，包含 331K 条多模态问答样本，用于对 Qwen3-VL 进行空间理解方面的监督微调；(iii) 开发 SpaitalAgent，这是一个集成 12 个专用空间感知工具的多智能体系统，支持 Plan-Execute 与 ReAct 两种推理范式，能够以免训练方式提升空间推理能力；(iv) 对 40 个代表性 MLLM 开展广泛评测，揭示了空间智能方面持续存在的挑战，同时证明了我们数据驱动与智能体方案的有效性。所有数据、代码与模型将公开可用。
英文原文摘要 1285 字符
Existing evaluations of multimodal large language models (MLLMs) on spatial intelligence are typically fragmented and limited in scope. In this work, we conduct a holistic assessment of the spatial understanding abilities of modern MLLMs and propose complementary data-driven and agent-based solutions. Concretely, we make the following contributions: (i) we propose SpatialScore, the most comprehensive and diverse multimodal spatial intelligence benchmark to date, encompassing various visual data types, input modalities, and question-answering formats with approximately 5K manually verified samples across 30 distinct tasks; (ii) we construct SpatialCorpus, a large-scale training resource with 331K multimodal QA samples for supervised fine-tuning Qwen3-VL on spatial understanding; (iii) we develop SpaitalAgent, a multi-agent system incorporating 12 specialized spatial perception tools, supporting both Plan-Execute and ReAct reasoning paradigms, enabling to improve spatial reasoning in a training-free manner; and (iv) we conduct extensive evaluations on 40 representative MLLMs, revealing persistent challenges in spatial intelligence while demonstrating the effectiveness of our data-driven and agent-based solutions. All data, code, and models will be publicly available.