Retrieval Feedback Memory Enhancement Large Model Retrieval Generation Method

TL;DR

This paper introduces RFM-RAG, a dynamic, feedback-driven retrieval method that enhances large language models by maintaining a continuous knowledge pool, leading to improved accuracy in question-answering tasks.

Contribution

It proposes a novel stateful retrieval framework with a feedback mechanism, transforming static retrieval into dynamic knowledge management for better LLM performance.

Findings

Outperforms previous RAG methods on three QA benchmarks.

Effectively maintains and updates an external evidence pool.

Improves system accuracy through iterative knowledge refinement.

Abstract

Large Language Models (LLMs) have shown remarkable capabilities across diverse tasks, yet they face inherent limitations such as constrained parametric knowledge and high retraining costs. Retrieval-Augmented Generation (RAG) augments the generation process by retrieving externally stored knowledge absent from the models internal parameters. However, RAG methods face challenges such as information loss and redundant retrievals during multi-round queries, accompanying the difficulties in precisely characterizing knowledge gaps for complex tasks. To address these problems, we propose Retrieval Feedback and Memory Retrieval Augmented Generation(RFM-RAG), which transforms the stateless retrieval of previous methods into stateful continuous knowledge management by constructing a dynamic evidence pool. Specifically, our method generates refined queries describing the models knowledge gaps using relational triples from questions and evidence from the dynamic evidence pool; Retrieves critical external knowledge to iteratively update this evidence pool; Employs a R-Feedback Model to evaluate evidence completeness until convergence. Compared to traditional RAG methods, our approach enables persistent storage of retrieved passages and effectively distills key information from passages to construct clearly new queries. Experiments on three public QA benchmarks demonstrate that RFM-RAG outperforms previous methods and improves overall system accuracy.