RAC: Efficient LLM Factuality Correction with Retrieval Augmentation

TL;DR

RAC is a retrieval-augmented post-correction method that improves the factual accuracy of large language models efficiently without additional fine-tuning, demonstrating significant performance gains.

Contribution

The paper introduces RAC, a novel retrieval-based post-correction technique that enhances LLM factuality with low latency and broad applicability without fine-tuning.

Findings

Achieves up to 30% improvement over baselines

Effective across multiple datasets and LLMs

Reduces latency compared to prior methods

Abstract

Large Language Models (LLMs) exhibit impressive results across a wide range of natural language processing (NLP) tasks, yet they can often produce factually incorrect outputs. This paper introduces a simple but effective low-latency post-correction method, \textbf{Retrieval Augmented Correction (RAC)}, aimed at enhancing the factual performance of LLMs without requiring additional fine-tuning. Our method is general and can be used with any instruction-tuned LLM, and has greatly reduced latency compared to prior approaches. RAC decomposes the LLM's output into atomic facts and applies a fine-grained verification and correction process with retrieved content to verify and correct the LLM-generated output. Our extensive experiments show that RAC yields up to 30\% improvements over state-of-the-art baselines across two popular factuality evaluation datasets, validating its efficacy and robustness in both with and without the integration of Retrieval-Augmented Generation (RAG) across different LLMs.\footnote{Our code is at \url{https://github.com/jlab-nlp/Retrieval-Augmented-Correction}}