Quantifying Factual Divergence in Generative Models: SHAP-LIME Based Hallucination Score for LLMs | Synapse