Spatial Reasoning
Spatial reasoning, the ability to understand and manipulate spatial relationships, is a crucial area of research in artificial intelligence, focusing on enabling machines to perform tasks requiring spatial understanding and manipulation. Current research emphasizes improving the spatial reasoning capabilities of large language models (LLMs) and vision-language models (VLMs) through techniques like prompt engineering, 3D scene graph integration, and the development of novel training datasets and benchmarks that specifically target spatial reasoning challenges. These advancements are significant because improved spatial reasoning is essential for progress in robotics, autonomous navigation, and other applications requiring interaction with the physical world.
Papers
Benchmarking Spatial Relationships in Text-to-Image Generation
Tejas Gokhale, Hamid Palangi, Besmira Nushi, Vibhav Vineet, Eric Horvitz, Ece Kamar, Chitta Baral, Yezhou Yang
Are Deep Neural Networks SMARTer than Second Graders?
Anoop Cherian, Kuan-Chuan Peng, Suhas Lohit, Kevin A. Smith, Joshua B. Tenenbaum