The field of 3D spatial reasoning is moving towards more explicit and generalizable representations, enabling machines to better understand and interact with complex environments. Recent work has focused on developing frameworks that can efficiently translate geometric data into actionable knowledge, and integrating spatial reasoning with machine learning, natural language processing, and rule systems. Notable papers include:

• Spatial Reasoner, which presents a framework for bridging geometric facts with symbolic predicates and relations, and demonstrates the capability to efficiently translate geometric data into actionable knowledge.
• SpaRE, which enhances spatial reasoning in vision-language models with synthetic data and shows strong improvements on spatial reasoning benchmarks.