How does Misinformation Affect Large Language Model Behaviors and Preferences?

TL;DR

This paper introduces MisBench, a large benchmark for evaluating how large language models handle misinformation, revealing their vulnerabilities and proposing a new method to improve misinformation detection.

Contribution

The paper presents MisBench, the largest misinformation benchmark for LLMs, and proposes Reconstruct to Discriminate (RtD), a novel approach to enhance misinformation detection capabilities.

Findings

LLMs are susceptible to knowledge conflicts and stylistic variations in misinformation.

MisBench effectively evaluates LLMs' behavior and knowledge preferences regarding misinformation.

RtD improves LLMs' ability to detect misinformation.

Abstract

Large Language Models (LLMs) have shown remarkable capabilities in knowledge-intensive tasks, while they remain vulnerable when encountering misinformation. Existing studies have explored the role of LLMs in combating misinformation, but there is still a lack of fine-grained analysis on the specific aspects and extent to which LLMs are influenced by misinformation. To bridge this gap, we present MisBench, the current largest and most comprehensive benchmark for evaluating LLMs' behavior and knowledge preference toward misinformation. MisBench consists of 10,346,712 pieces of misinformation, which uniquely considers both knowledge-based conflicts and stylistic variations in misinformation. Empirical results reveal that while LLMs demonstrate comparable abilities in discerning misinformation, they still remain susceptible to knowledge conflicts and stylistic variations. Based on these findings, we further propose a novel approach called Reconstruct to Discriminate (RtD) to strengthen LLMs' ability to detect misinformation. Our study provides valuable insights into LLMs' interactions with misinformation, and we believe MisBench can serve as an effective benchmark for evaluating LLM-based detectors and enhancing their reliability in real-world applications. Codes and data are available at https://github.com/GKNL/MisBench.