This research introduces a combined detection and mitigation framework to address the critical problem of hallucinations in healthcare LLMs specifically for Italian medical applications.

• Combines Retrieval-Augmented Generation (RAG) for hallucination mitigation by grounding responses in external data
• Introduces the novel Negative Missing Information Scoring System (NMISS) to detect hallucinations based on contextual relevance
• Demonstrates improved performance in real-world Italian healthcare question-answering tasks
• Provides a practical framework that can help maintain factual accuracy in sensitive medical contexts

Business Impact: This approach helps build more trustworthy AI systems for healthcare, reducing potential liability and improving patient safety when deploying LLM-based medical assistants.

Original Paper: Addressing Hallucinations with RAG and NMISS in Italian Healthcare LLM Chatbots