核心概念
大規模言語モデルのゼロショットパフォーマンスとプロンプト戦略の効果を評価する。
引用
"Task-specific fine-tuned models generally tend to outperform LLMs in zero-shot settings."
"More detailed and complex prompts do not necessarily enhance classification performance."
"LLMs can potentially be used for data annotation as a way to reduce human annotation costs."