감정적 표현이 AI 성능을 바꿀까? — 프롬프트 감정 프레이밍의 실제 효과

Despite widespread belief in emotional prompting, this research shows that simple emotional framing has negligible effect on LLM performance across standard benchmarks, contradicting anecdotal success stories. The breakthrough is that adaptive emotion selection—where an agent dynamically chooses emotions based on input—consistently improves results, suggesting that emotional framing works best as a personalized routing strategy rather than a generic technique. This shifts the paradigm from static emotional templates to intelligent, task-aware prompt optimization.