Can You Answer This? - Exploring Zero-Shot QA Generalization Capabilities in Large Language Models