Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unmasking Digital Falsehoods: A Comparative Analysis of LLM-Based Misinformation Detection Strategies

Tianyi Huang, Jingyuan Yi, Peiyang Yu, Xiaochuan Xu

TL;DR

The paper addresses the challenge of detecting misinformation on social media using Large Language Models (LLMs). It presents a comparative analysis of text-based, multimodal, and agentic versus non-agentic detection strategies, evaluating fine-tuned and zero-shot approaches across domains such as politics, health, and finance, using benchmarks like FakeNewsNet, SciNews, MM-COVID, and LIAR. Key contributions include a taxonomy of misinformation, a performance-oriented comparison across detection pipelines, and a discussion of explainability and evaluation challenges, with insights into the trade-offs between accuracy, speed, and transparency. The findings advocate hybrid architectures that combine structured verification with adaptive learning to improve detection accuracy and explainability, and outline practical directions for real-time tracking, federated learning, and cross-platform detection.

Abstract

The proliferation of misinformation on social media has raised significant societal concerns, necessitating robust detection mechanisms. Large Language Models such as GPT-4 and LLaMA2 have been envisioned as possible tools for detecting misinformation based on their advanced natural language understanding and reasoning capabilities. This paper conducts a comparison of LLM-based approaches to detecting misinformation between text-based, multimodal, and agentic approaches. We evaluate the effectiveness of fine-tuned models, zero-shot learning, and systematic fact-checking mechanisms in detecting misinformation across different topic domains like public health, politics, and finance. We also discuss scalability, generalizability, and explainability of the models and recognize key challenges such as hallucination, adversarial attacks on misinformation, and computational resources. Our findings point towards the importance of hybrid approaches that pair structured verification protocols with adaptive learning techniques to enhance detection accuracy and explainability. The paper closes by suggesting potential avenues of future work, including real-time tracking of misinformation, federated learning, and cross-platform detection models.

Unmasking Digital Falsehoods: A Comparative Analysis of LLM-Based Misinformation Detection Strategies

TL;DR

The paper addresses the challenge of detecting misinformation on social media using Large Language Models (LLMs). It presents a comparative analysis of text-based, multimodal, and agentic versus non-agentic detection strategies, evaluating fine-tuned and zero-shot approaches across domains such as politics, health, and finance, using benchmarks like FakeNewsNet, SciNews, MM-COVID, and LIAR. Key contributions include a taxonomy of misinformation, a performance-oriented comparison across detection pipelines, and a discussion of explainability and evaluation challenges, with insights into the trade-offs between accuracy, speed, and transparency. The findings advocate hybrid architectures that combine structured verification with adaptive learning to improve detection accuracy and explainability, and outline practical directions for real-time tracking, federated learning, and cross-platform detection.

Abstract

The proliferation of misinformation on social media has raised significant societal concerns, necessitating robust detection mechanisms. Large Language Models such as GPT-4 and LLaMA2 have been envisioned as possible tools for detecting misinformation based on their advanced natural language understanding and reasoning capabilities. This paper conducts a comparison of LLM-based approaches to detecting misinformation between text-based, multimodal, and agentic approaches. We evaluate the effectiveness of fine-tuned models, zero-shot learning, and systematic fact-checking mechanisms in detecting misinformation across different topic domains like public health, politics, and finance. We also discuss scalability, generalizability, and explainability of the models and recognize key challenges such as hallucination, adversarial attacks on misinformation, and computational resources. Our findings point towards the importance of hybrid approaches that pair structured verification protocols with adaptive learning techniques to enhance detection accuracy and explainability. The paper closes by suggesting potential avenues of future work, including real-time tracking of misinformation, federated learning, and cross-platform detection models.

Paper Structure

This paper contains 26 sections, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Background and Motivation
  3. Research Objectives
  4. Contributions of This Review
  5. Taxonomy of Misinformation and Detection Challenges
  6. Types of Misinformation
  7. Characteristics of Misinformation
  8. Challenges in Misinformation Detection
  9. Comparison of Detection Challenges Across Approaches
  10. Comparative Analysis of LLM-Based Misinformation Detection Methods
  11. Text-Based Approaches
  12. Hybrid Attention Framework: Integrating Statistical and Semantic Features
  13. Multimodal Approaches
  14. Agentic vs. Non-Agentic Models
  15. Comparison of Detection Pipelines
  16. ...and 11 more sections

Figures (2)

  • Figure 1: Reference from Qi, Peng, et al. "SNIFFER: Multimodal Large Language Model for Explainable Out-of-Context Misinformation Detection." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2024.
  • Figure 2: Reference from Tahmasebi, Sahar, Eric Müller-Budack, and Ralph Ewerth. "Multimodal misinformation detection using large vision-language models." Proceedings of the 33rd ACM International Conference on Information and Knowledge Management. 2024.