The ability of large language models (LLMs) to generate factually correct output remains relatively unexplored due to the lack of fact-checking and knowledge grounding during training and inference. In this work, we aim to address this challenge through the →