Spatial reasoning, the ability to represent and reason about spatial relations, geometry, topology, and embodied environments, is increasingly vital for evaluating and deploying large language models (LLMs) and vision–language models (VLMs). While multimodal foundation models show strong performance on spatial benchmarks, navigation, and embodied planning, much of this stems from shallow pattern matching rather than grounded spatial understanding. This review offers a comprehensive, framework-driven analysis of spatial reasoning in LLMs and VLMs, covering tasks, representations, datasets, benchmarks, architectures, training, and failure modes. We also link spatial reasoning to embodied AI, emphasizing challenges in generalization, compositionality, and grounding. Our survey unifies diverse fields, NLP, vision, robotics, and cognitive science, to guide the development of truly spatially grounded foundation models.

This paper investigates the potential of using natural language descriptions as an alternative to direct image-based observations for learning policies in reinforcement learning. Due to the inherent challenges in managing image-based observations, which include abundant information and irrelevant features, we propose a method that compresses images into a natural language form for state representation. This approach allows better interpretability and leverages the processing capabilities of large-language models. We conducted several experiments involving tasks that required image-based observation. The results demonstrated that policies trained using natural language descriptions of images yield better generalization than those trained directly from images, emphasizing the potential of this approach in practical settings.