评测视觉语言模型在多实例空间推理中的几何精度，提出无需训练的新框架。
SpatialQuery: Benchmarking Geometry-Grounded Multi-Instance Spatial Reasoning in Vision-Language Models
- 通过场景立方化将图像转为鸟瞰图，用统一方块表示物体相对位置。
- 在100万+问题上测试，距离估计误差仅0.259米，近距判断准确率达84.18%。
- 适合关注视觉定位、空间理解的开发者和研究者使用。
视觉语言模型（VLMs）在语义理解上表现优异，但在度量空间推理方面仍不可靠，尤其在需要比较同一类别多个实例时。本文通过最近实例距离查询（CIDQ）研究该问题，要求模型识别与唯一参考对象最近的可见候选物，并估算其重力对齐的地平面距离。我们提出SPATIALQUERY，一个无需训练的单帧RGB图像推理框架，配套构建了包含超过一百万条仅基于RGB的问答对的SPATIALQUERY-1M基准数据集，来自200个室内场景。SPATIALQUERY通过场景立方化恢复实例级度量几何，并转换为规范化的鸟瞰视图，将物体表示为大小一致、按类别编码的方块，以强调其相对地平面位置。我们进一步提出不确定性感知链式思维（UA-CoT）提示方法，将几何推导出的每实例不确定性引入VLM推理过程。在不进行任务特定微调或架构修改的前提下，结合Qwen3-VL-8B模型，SPATIALQUERY实现了0.259米的地板均方误差（Floor-MAE）、90.5%的0.3米内准确率（[email protected] m）以及84.18%的邻近决策准确率，优于微调过的空间专用模型、通用型VLM及闭源前沿模型。代码、基准资源与交互演示可访问 https://namhai1810.github.io/SpatialQuery/。
原文摘要 · Abstract (English)
Vision-language models (VLMs) achieve strong semantic understanding but remain unreliable in metric spatial reasoning, particularly when queries require comparing multiple instances of the same object category. We study this problem through the Closest-Instance Distance Query (CIDQ), where a model must identify the nearest visible candidate to a unique reference object and estimate their gravity-aligned floor-plane distance. We introduce SPATIALQUERY, a training- free framework for CIDQ reasoning from a single RGB image, together with SPATIALQUERY-1M, a benchmark containing over one million RGB-only question-answer pairs from 200 indoor scenes. SPATIALQUERY recovers instance-level metric geometry and transforms it into a canonical Bird's-Eye View through Scene Cubifying, which represents objects as uniformly sized, category-coded blocks to emphasize their relative floor- plane locations. We further propose Uncertainty-Aware Chain-of-Thought (UA-CoT) prompting, which incorporates geometry- derived per-instance uncertainty into the VLM reasoning process. Without task-specific fine-tuning or architectural modification, SPATIALQUERY with Qwen3-VL-8B achieves a Floor-MAE of 0.259 m, an [email protected] m of 90.5%, and a proximity-decision accuracy of 84.18%, outperforming fine-tuned spatial specialists, general-purpose VLMs, and closed-source frontier models. Code, benchmark resources, and an interactive demo are available at https://namhai1810.github.io/SpatialQuery/.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。