Not All RAGs Are Created Equal: A Component-Wise Empirical Study for Software Engineering Tasks

Qiang Ke, Yanjie Zhao, Hongjin Leng +2

While Retrieval-Augmented Generation (RAG) is increasingly adopted to ground Large Language Models (LLMs) in software artifacts, the optimal configuration of its components remains…