Search papers, labs, and topics across Lattice.
This paper investigates the spatial numerical understanding of Vision-Language Models (VLMs) through a novel framework called SpaceNum, which evaluates how well these models can connect numerical outputs to spatial perception. By formulating two bidirectional tasks, Num2Space and Space2Num, the authors reveal that current VLMs often fail to ground numerical values in meaningful spatial contexts, performing close to random guessing in many scenarios. The study highlights significant limitations in VLMs' ability to construct stable, coordinate-aware representations and suggests that while tuning can yield some improvements, explicit reasoning offers only marginal benefits.
VLMs struggle to meaningfully ground numerical outputs in spatial contexts, often performing at chance levels in critical tasks.
Vision-Language Models (VLMs) are increasingly deployed in embodied environments, where they need produce numerical outputs such as action magnitudes and spatial coordinates. Although these numbers appear meaningful, it remains unclear whether these numerical outputs are genuinely grounded in spatial perception. Therefore, in this work, we revisit spatial numerical understanding through SpaceNum, a unified framework that captures two complementary settings: numbers as dynamic transitions during spatial exploration, and numbers as static layouts in spatial reasoning. We formulate two bidirectional tasks, Num2Space and Space2Num, to evaluate how well VLMs map between vision-side spatial structure and language-side numerical representations. We systematically study whether current VLMs truly understand numerical values in spatial settings. Across dynamic transitions and static layouts, we find that models largely fail to ground numbers in spatial meaning and often perform close to random guess. Through error analysis, reasoning trace analysis, and controlled interventions, we show that current VLMs rely heavily on shallow spatial cues, struggle to build stable coordinate-aware representations, and fail to abstract structured spatial layouts from visual observations. We further show that explicit reasoning provides only marginal gains, while tuning can partially improve spatial numerical understanding and transfer to external spatial reasoning benchmarks.
