想象感知标记增强多模态语言模型的空间推理能力
Imaginative Perception Tokens Enhance Spatial Reasoning in Multimodal Language Models
AI总结：
提出想象感知标记（IPT）作为中间感知表征，通过监督学习提升多模态语言模型在不可见视角推理、遮挡路径追踪等空间推理任务上的性能，在三个新构建的数据集上优于文本思维链训练。
摘要：
视觉语言模型（VLM）在许多任务上表现出色，但当关键信息无法直接观察时，仍难以进行空间推理。许多此类问题需要想象感知：从未见视角推断所见内容、追踪穿过遮挡空间的路径、或将部分观察整合成连贯的空间表征。我们引入了想象感知标记（IPT），这是一种中间感知表征，将VLM在替代空间配置下会感知到的内容外部化，同时保持与观察输入一致。为了研究这一能力，我们设计了三个任务：视角推理（PET）、路径追踪（PT）和多视角计数（MVC），并构建了包含约20K个样本的数据集，附带真实想象、答案和评估基准。以统一VLM BAGEL为骨干，IPT监督持续提升了空间推理性能，并且通常优于文本思维链训练，即使在推理时不生成图像。在MVC上，IPT将准确率提高了3.4%，并在PT上达到了与强大闭源模型竞争的性能。我们进一步发现，将IPT与仅标签监督相结合能带来额外收益，而文本思维链可能大幅降低性能，这表明当空间计算被迫通过语言进行时存在模态不匹配。总体而言，IPT为推理未观察到的空间结构提供了原则性的监督信号，在生成可解释中间表征的同时提升了泛化能力。
英文摘要：
Vision language models (VLMs) excel at many tasks but still struggle with spatial reasoning when critical information is not directly observable. Many such problems require imaginative perception: inferring what would be seen from an unseen viewpoint, tracing paths through occluded spaces, or integrating partial observations into a coherent spatial representation. We introduce Imaginative Perception Tokens (IPT), intermediate perceptual representations that externalize what a VLM would perceive under alternative spatial configurations while remaining consistent with the observed input. To study this capability, we formulate three tasks, Perspective Taking (PET), Path Tracing (PT), and Multiview Counting (MVC), and construct datasets of approximately 20K examples with ground truth imaginations, answers, and evaluation benchmarks. Using the unified VLM BAGEL as the backbone, IPT supervision consistently improves spatial reasoning and often outperforms textual chain of thought training, even without generating images at inference time. On MVC, IPT improves accuracy by 3.4% and achieves competitive performance with strong closed-source models on PT. We further find that combining IPT and label-only supervision yields additional gains, whereas textual chain of thought can substantially degrade performance, suggesting a modality mismatch when spatial computation is forced through language. Overall, IPT provides a principled supervision signal for reasoning about unobserved spatial structure, improving generalization while producing interpretable intermediate representations.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
