試験
通常
試験時間: 00:00問題時間: 00:00
問題一覧
問題 64 / 65
I want to directly manipulate the model’s abstract internal representations to experimentally alter its behavior. Which options involve its internal latent representations rather than user-level prompt techniques?
①Give instructions in the instruction prompt.
②Specify exclusions in the negative prompt.
③Example using a few-shot prompt.
④Assigning a role in the system prompt
⑤Manipulation of the model's latent space