試験
通常
試験時間: 00:00問題時間: 00:00
問題一覧
問題 11 / 65
I want to reduce misinformation (hallucinations) in a legal-advice chatbot and provide answers supported by evidence. Which method is most effective?
①Upgrade to a larger general-purpose language model.
②Operate solely by directly fine-tuning on user data.
③Completely replace it with a rule-based system.
④Introduce RAG (retrieval-augmented generation) that references an external knowledge base.
⑤Add random noise to the generated outputs to increase diversity.