From USD Scenes to Knowledge Graphs: Zero-Shot Ontology Grounding with LLMs
Researchers have developed a method using large language models (LLMs) to automatically ground objects in 3D simulation scenes to formal ontology classes. This approach aims to overcome the limitations of manually curated dictionaries, which are often brittle and lack generalization. The LLMs demonstrated high accuracy in mapping scene objects to ontology classes, significantly outperforming traditional baselines, especially when provided with contextual cues from the scene graph. AI
IMPACT Automates a key step in robot reasoning by enabling LLMs to interpret 3D simulation environments.