BlendFilter: Advancing Retrieval-Augmented Large Language Models via Query Generation Blending and Knowledge Filtering

TL;DR

BlendFilter enhances retrieval-augmented LLMs by combining query generation blending with knowledge filtering, effectively improving performance on knowledge-intensive tasks despite noisy retrieval issues.

Contribution

The paper introduces BlendFilter, a novel method that integrates query blending and knowledge filtering to improve retrieval-augmented LLM performance.

Findings

Outperforms state-of-the-art baselines on three QA benchmarks

Effectively reduces noise in retrieved knowledge

Enhances model accuracy in complex queries

Abstract

Retrieval-augmented Large Language Models (LLMs) offer substantial benefits in enhancing performance across knowledge-intensive scenarios. However, these methods often face challenges with complex inputs and encounter difficulties due to noisy knowledge retrieval, notably hindering model effectiveness. To address this issue, we introduce BlendFilter, a novel approach that elevates retrieval-augmented LLMs by integrating query generation blending with knowledge filtering. BlendFilter proposes the blending process through its query generation method, which integrates both external and internal knowledge augmentation with the original query, ensuring comprehensive information gathering. Additionally, our distinctive knowledge filtering module capitalizes on the intrinsic capabilities of the LLM, effectively eliminating extraneous data. We conduct extensive experiments on three open-domain question answering benchmarks, and the findings clearly indicate that our innovative BlendFilter surpasses state-of-the-art baselines significantly.