EAReranker: Efficient Embedding Adequacy Assessment for Retrieval Augmented Generation
With the increasing adoption of Retrieval-Augmented Generation (RAG) systems for knowledge-intensive tasks, ensuring the adequacy of retrieved documents has become critically important for generation quality. Traditional reranking approaches face three significant challenges: substantial computation…