Towards Lightweight Reliability: Using Soft Prompts for Hallucination Mitigation in Large Language Models | ArxivCSExplorer