S-Space:探索多模态语言模型中的空间工作区
原标题:Paper page - S-Space: Exploring Spatial Workspace in Multimodal Language Models
AI 摘要
该论文提出 S-Space,即多模态语言模型(MLLMs)内部存在的一个线性表示子空间,物体 token 的表示在其中编码了水平、垂直和距离三个连续坐标轴。研究发现 S-Space 是模型内在的而非由特定数据或提示诱导,它作为内部空间工作区,支持感知编码位置、内部推理变换视角以及语言回答的选择性读取。通过因果干预和案例研究,作者还将其用作诊断工具,揭示 MLLMs 能形成空间表示但在跨视角变换上存在瓶颈。
正文节选
S-Space: Exploring Spatial Workspace in Multimodal Language Models Abstract Multimodal language models (MLLMs) exhibit promising spatial reasoning capabilities, yet the internal mechanisms supporting them remain poorly understood. We uncover S-Space, a linear representation subspace in which object-token representations encode continuous coordinates along horizontal, vertical, and distance axes. Although we identify these axes by differentiating spatial-answer scores with respect to object-token