AI 中文总结
本文扩展ToS框架为EMRD流程，评估VLMs在安全关键场景下的空间理解与决策能力，发现其决策依赖文本先验、空间推理受低光照影响，且记忆与人类认知存在根本差异，存在对齐风险。
AI 中文摘要
空间理论框架（Theory of Space framework，简称ToS）用于评估好奇心驱动的视觉语言模型（Vision-Language Models，简称VLMs）在部分可观测性下的空间理解能力。随着AI技术越来越多地应用于安全关键场景，理解VLMs是否具备稳健的空间记忆并做出可靠决策至关重要。在本文中，我们评估VLMs的决策是基于物理证据还是受视觉语言偏差影响，其记忆过程是否符合人类认知模式，以及它们如何应对环境危害。我们将ToS框架扩展为安全关键、目标驱动的流程，命名为探索、建图、记忆、决策（Explore, Map, Remember, and Decide，简称EMRD）。随后，我们通过环境覆盖度和时间效率指标量化探索能力（Explore），评估空间保真度（Map），并通过一套心理指标评估记忆持久性（Remember），再利用焦点指标衡量认知决策能力（Decide）。我们的结果显示，在决策能力方面，VLMs经常基于预训练文本先验选择疏散点，却缺乏支撑其选择的空间依据；空间推理在低光照条件下会下降，但不受纹理和颜色篡改的影响。我们的发现表明，VLM记忆与人类认知存在根本差异，会产生不可预测的对齐风险。
英文摘要
Theory of Space framework (ToS) assesses the spatial understanding of curiosity-driven Vision-Language Models (VLMs) under partial observability. As AI techniques are increasingly applied to safety-critical scenarios, it is crucial to understand whether VLMs possess robust spatial memory and make reliable decisions. In this paper, we assess whether VLMs' decisions are based on physical evidence or are corrupted by visual-language biases, if their memory processes align with human cognitive patterns, and how they respond to environmental hazards. We extend the ToS framework into a safety-critical, goal-driven pipeline, named Explore, Map, Remember, and Decide (EMRD). We then quantify Exploration Competence (Explore) through metrics of environmental coverage and temporal efficiency, assess Spatial Fidelity (Map), evaluate, with a suite of psychological metrics, Memory Persistence (Remember), and measure, using focal-point metrics, Cognitive Decision-Making (Decide). Our results show that in terms of decision-making capabilities, VLMs frequently select evacuation points based on pre-trained textual priors while lacking the spatial grounding to justify their choices. We also show that spatial reasoning degrades in low-light conditions, but it is not affected by texture and colour tampering. Our findings suggest that VLM memory fundamentally diverges from human cognition, creating unpredictable risks of misalignment.
