“In-context hallucination in large language models occurs when the model's output is inconsistent with the provided source content.”