ZoFia: Zero-Shot Fake News Detection with Entity-Guided Retrieval and Multi-LLM Interaction

2025-11-02

AI Summary

Falsehoods in news propagate rapidly, threatening social stability; however, existing detection methods exhibit poor generalization to emerging topics—static models rely heavily on historical data, while large language models (LLMs) suffer from knowledge obsolescence and hallucination risks. This paper proposes a zero-shot, training-free two-stage framework: (1) a hierarchical entity saliency modeling module coupled with SC-MMR keyword selection for precise retrieval; and (2) a multi-role LLM collaborative debate mechanism integrating adversarial reasoning and interpretable analysis. The method requires no labeled data or fine-tuning, significantly enhancing robustness against previously unseen news topics. Evaluated on two public benchmarks, it outperforms state-of-the-art zero-shot baselines and most few-shot approaches, demonstrating superior generalizability and practical deployability.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Machine Learning: Large Multimodal Models (LMMs)
• Application Domains: Misinformation & Fake News

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

The rapid spread of fake news threatens social stability and public trust, rendering its detection an imperative research priority. Although large language models (LLMs) excel at numerous natural language processing tasks with their remarkable contextual understanding and extensive prior knowledge, the time-bounded knowledge coverage and tendency for generating hallucination content reduce their reliability when handling fast-evolving news streams. Furthermore, models trained on existing static datasets also often lack the generalization needed for emerging news topics. To address these challenges, we propose ZoFia, a novel two-stage zero-shot fake news detection framework. First, we introduce Hierarchical Salience to quantify the importance of entities in the news content, and propose the SC-MMR algorithm to effectively select an informative and diverse set of keywords that serve as queries for retrieving up-to-date external evidence. Subsequently, a multi LLM interactive system, in which each agent assumes a distinct role, performs multi-view collaborative analysis and adversarial debate over the news text and its related information, and finally produces an interpretable and robust judgment. Comprehensive experiments on two public datasets demonstrate that ZoFia obviously outperforms existing zero-shot baselines and most of few-shot methods. Our codes will be open-sourced to facilitate related communities.

Problem

• Detecting fake news in rapidly evolving news streams
• Overcoming LLM limitations like outdated knowledge and hallucinations
• Improving generalization for emerging news topics without training data

Innovation

• Entity-guided retrieval using hierarchical salience quantification
• Multi-LLM interactive system with role-based agents
• Adversarial debate for multi-view collaborative analysis