A South Korean team has found a way to mitigate hallucinations in LLMs