試験
通常
試験時間: 00:00問題時間: 00:00
問題一覧
問題 24 / 65
What is the most effective measure to prevent malicious external input (prompt injection) while clearly conveying operational rules to the model?
①Just keep the output short.
②Teach by providing many examples.
③Lower the temperature and top_p.
④Add random noise in advance.
⑤Give strict instructions in the system role and combine them with input validation.