To minimise hallucinations and enhance the efficiency of RAG-powered Question Answering systems, with OpenAI log probabilities can be used to assess the model’s confidence in its generated content bas