Prompt caching으로 긴 문맥 LLM의 self-consistency 비용 절감

Prompt caching makes self-consistency cheap for long-context LLMs