The First Workshop on Efficient Spatial Reasoning

Spatial reasoning—the ability to understand and manipulate spatial relations among objects and environments—is essential for both human cognition and AI applications such as robotics, autonomous navigation, and 3D perception. While traditional geometric and symbolic methods have enabled structured spatial analysis, modern foundation models embed spatial reasoning within large-scale learning systems, yielding impressive results in tasks like 3D reconstruction and multimodal reasoning. However, these models often rely on large parameters and inefficient inference strategies, leading to poor performance on complex, multi-step spatial reasoning tasks. Recent advances in efficient reasoning aim to improve interpretability and reduce computational cost, yet they rarely address the unique robustness and safety demands of spatial reasoning in embodied and real-world settings. To bridge this gap, the proposed workshop will unite researchers and practitioners to explore adaptive, efficient, and reliable spatial reasoning for foundation models, fostering dialogue across AI, cognitive science, and applied domains.

Goals of our workshop:

• Bringing together researchers working on efficient spatial reasoning.

•  Foster interdisciplinary collaboration and facilitate the exchange of ideas across traditionally siloed research communities. 

Organizers

Acknowledgement