LLM-Independent Adaptive RAG: Let the Question Speak for Itself

TL;DR

This paper introduces lightweight, LLM-independent adaptive retrieval methods for question answering that utilize external information, achieving comparable performance to complex LLM-based methods with improved efficiency.

Contribution

The study proposes novel external information-based adaptive retrieval techniques that do not rely on LLM uncertainty, enhancing efficiency while maintaining QA performance.

Findings

Achieves similar QA performance as complex LLM-based methods.

Significantly improves retrieval efficiency.

Effective use of external features for adaptive retrieval.

Abstract

Large Language Models~(LLMs) are prone to hallucinations, and Retrieval-Augmented Generation (RAG) helps mitigate this, but at a high computational cost while risking misinformation. Adaptive retrieval aims to retrieve only when necessary, but existing approaches rely on LLM-based uncertainty estimation, which remain inefficient and impractical. In this study, we introduce lightweight LLM-independent adaptive retrieval methods based on external information. We investigated 27 features, organized into 7 groups, and their hybrid combinations. We evaluated these methods on 6 QA datasets, assessing the QA performance and efficiency. The results show that our approach matches the performance of complex LLM-based methods while achieving significant efficiency gains, demonstrating the potential of external information for adaptive retrieval.