Exploring the Generalizability of Factual Hallucination Mitigation via Enhancing Precise Knowledge Utilization

TL;DR

This paper introduces PKUE, a method to improve large language models' ability to accurately utilize knowledge, reducing factual hallucinations and enhancing performance across diverse factual and general tasks.

Contribution

The paper proposes PKUE, a novel fine-tuning approach that enhances LLMs' knowledge utilization by training on self-generated responses to factual questions.

Findings

PKUE significantly improves factual accuracy in LLMs.

Enhanced performance across multiple languages and task types.

FactualBench dataset enables comprehensive evaluation.

Abstract

Large Language Models (LLMs) often struggle to align their responses with objective facts, resulting in the issue of factual hallucinations, which can be difficult to detect and mislead users without relevant knowledge. Although post-training techniques have been employed to mitigate the issue, existing methods usually suffer from poor generalization and trade-offs in other different capabilities. In this paper, we propose to address these by directly augmenting LLM's fundamental ability to precisely leverage its knowledge and introduce PKUE (Precise Knowledge Utilization Enhancement), which fine-tunes the model on self-generated responses to precise and simple factual questions through preference optimization. Furthermore, we construct FactualBench, a comprehensive and precise factual QA dataset containing 181k Chinese data spanning 21 domains, to facilitate both evaluation and training. Extensive experiments demonstrate that PKUE significantly improves LLM overall performance, with consistent enhancement across factual tasks of various forms, general tasks beyond factuality, and tasks in different language.