Optimizing Knowledge Integration in Retrieval-Augmented Generation with Self-Selection

TL;DR

This paper introduces a Self-Selection RAG framework that improves knowledge integration in LLMs by enabling the model to select the best response from internally generated and externally retrieved knowledge, enhancing accuracy.

Contribution

The paper presents a novel Self-Selection RAG method with a training approach using DPO on RGP data, improving knowledge integration in LLMs.

Findings

Outperforms baseline methods on Natural Questions and TrivialQA datasets.

Effective in enhancing accuracy by combining internal and external knowledge.

Demonstrates superiority with open-source LLMs like Llama2-13B-Chat and Mistral-7B.

Abstract

Retrieval-Augmented Generation (RAG), which integrates external knowledge into Large Language Models (LLMs), has proven effective in enabling LLMs to produce more accurate and reliable responses. However, it remains a significant challenge how to effectively integrate external retrieved knowledge with internal parametric knowledge in LLMs. In this work, we propose a novel Self-Selection RAG framework, where the LLM is made to select from pairwise responses generated with internal parametric knowledge solely and with external retrieved knowledge together to achieve enhanced accuracy. To this end, we devise a Self-Selection-RGP method to enhance the capabilities of the LLM in both generating and selecting the correct answer, by training the LLM with Direct Preference Optimization (DPO) over a curated Retrieval Generation Preference (RGP) dataset. Experimental results with two open-source LLMs (i.e., Llama2-13B-Chat and Mistral-7B) well demonstrate the superiority of our approach over other baseline methods on Natural Questions (NQ) and TrivialQA datasets.