(논문 요약) Speculative RAG: Enhancing Retrieval Augmented Generation through Drafting (Paper)
핵심 내용
- specialists RAG drafter 로 여러 관점에서 생성뒤, general LM 으로 평가해서 best 선택
- algorithm
실험 결과
- increase accuracy by up to 12.97%, reduce latency by 51% compared to conventional RAG systems on PubHealth