Cognitive-Aligned Document Selection for Retrieval-augmented Generation

TL;DR

This paper introduces GGatrieval, a method that dynamically refines queries and filters documents in retrieval-augmented generation to improve the factual accuracy and verifiability of large language model outputs.

Contribution

The paper presents a novel grounded alignment and dynamic semantic compensation mechanism for better document retrieval in RAG systems, enhancing factual correctness.

Findings

Achieves state-of-the-art results on the ALCE benchmark.

Significantly improves the supportiveness of retrieved documents.

Enhances the factual accuracy of generated responses.

Abstract

Large language models (LLMs) inherently display hallucinations since the precision of generated texts cannot be guaranteed purely by the parametric knowledge they include. Although retrieval-augmented generation (RAG) systems enhance the accuracy and reliability of generative models by incorporating external documents, these retrieved documents often fail to adequately support the model's responses in practical applications. To address this issue, we propose GGatrieval (Fine-\textbf{G}rained \textbf{G}rounded \textbf{A}lignment Re\textbf{trieval} for verifiable generation), which leverages an LLM to dynamically update queries and filter high-quality, reliable retrieval documents. Specifically, we parse the user query into its syntactic components and perform fine-grained grounded alignment with the retrieved documents. For query components that cannot be individually aligned, we propose a dynamic semantic compensation mechanism that iteratively refines and rewrites the query while continuously updating the retrieval results. This iterative process continues until the retrieved documents sufficiently support the query's response. Our approach introduces a novel criterion for filtering retrieved documents, closely emulating human strategies for acquiring targeted information. This ensures that the retrieved content effectively supports and verifies the generated outputs. On the ALCE benchmark, our method significantly surpasses a wide range of baselines, achieving state-of-the-art performance.