A recent study evaluates large language models (LLMs) for goal recognition tasks, revealing varied performance based on evidence integration rather than just world knowledge. This positions goal recognition as a valuable benchmark for assessing LLMs' foundational planning capabilities.