The content introduces the Adaptive-RAG framework for adapting retrieval-augmented Large Language Models (LLMs) based on query complexity. It addresses the limitations of existing approaches by dynamically selecting strategies ranging from non-retrieval to multi-step retrieval based on query complexities. The framework includes a classifier to predict query complexity levels and automatically collects training data without human labeling. Experimental results demonstrate improved efficiency and accuracy compared to traditional adaptive retrieval strategies.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Soyeong Jeon... at arxiv.org 03-22-2024
https://arxiv.org/pdf/2403.14403.pdfDeeper Inquiries