【論文要約:自動運転関連】LLMs Can Check Their Own Results to Mitigate Hallucinations in Traffic Understanding Tasks