Reasoning through Evolution: Automatic Meta-path Discovery for LLM-based Fake News Detection

ACL ARR 2026 January Submission8425 Authors

Keywords: Fake News Detection; Large Language Models; Graph Reasoning

Abstract:

Propagation structures provide crucial evidence for fake news detection, yet existing approaches primarily rely on supervised GNN-based models, which require substantial labeled data and exhibit limited generalization. Although large language models (LLMs) exhibit strong reasoning capabilities, they struggle to directly exploit propagation graphs and remain unreliable for zero-shot and few-shot fake news detection. To bridge this gap, we propose MAGER, a multi-agent genetic evolution framework
for meta-path discovery with reasoning guidance, enabling frozen LLMs to incorporate propagation structures for veracity assessment.
We further introduce a graph in-context learning strategy to retrieve both semantic and structurally similar demonstrations to enhance LLM's classification and reasoning ability.
Extensive experiments show that MAGER significantly enhances LLMs as standalone fake news detectors in a data-efficient setting.

Paper Type: Long

Research Area: NLP Applications

Research Area Keywords: rumor/misinformation detection

Contribution Types: NLP engineering experiment

Languages Studied: English

Submission Number: 8425