Ahead of the Spread: Agent-Driven Virtual Propagation for Early Fake News Detection

TL;DR

The paper tackles the challenge of detecting fake news at the very early stage when real propagation signals are scarce. It introduces AVOID, an agent-driven framework where LLM-powered Diffuser and Verifier agents, grounded in data-driven personas, simulate plausible virtual propagation to supply auxiliary diffusion evidence alongside content features. A key contribution is the denoising-guided fusion that aligns content and simulated propagation using a variational approach with a symmetric KL regularizer, enabling robust early detection without real diffusion data. Experiments on real-world datasets show that AVOID consistently outperforms state-of-the-art baselines, with strong performance in early-stage scenarios and compelling evidence that virtual propagation can meaningfully supplement content-based signals for timely misinformation detection.

Abstract

Early detection of fake news is critical for mitigating its rapid dissemination on social media, which can severely undermine public trust and social stability. Recent advancements show that incorporating propagation dynamics can significantly enhance detection performance compared to previous content-only approaches. However, this remains challenging at early stages due to the absence of observable propagation signals. To address this limitation, we propose AVOID, an \underline{a}gent-driven \underline{v}irtual pr\underline{o}pagat\underline{i}on for early fake news \underline{d}etection. AVOID reformulates early detection as a new paradigm of evidence generation, where propagation signals are actively simulated rather than passively observed. Leveraging LLM-powered agents with differentiated roles and data-driven personas, AVOID realistically constructs early-stage diffusion behaviors without requiring real propagation data. The resulting virtual trajectories provide complementary social evidence that enriches content-based detection, while a denoising-guided fusion strategy aligns simulated propagation with content semantics. Extensive experiments on benchmark datasets demonstrate that AVOID consistently outperforms state-of-the-art baselines, highlighting the effectiveness and practical value of virtual propagation augmentation for early fake news detection. The code and data are available at https://github.com/Ironychen/AVOID.

Figures (7)

• Figure 1: Figures (a)-(c) compare the architectures of different fake news detection methodologies, while Figure (d) presents the comparative performance across two datasets, confirming the significant enhancement provided by propagation-enhanced methods over content-only baselines.
• Figure 2: Overview of AVOID. Real-world comments are distilled into fine-grained personas, which initialize simulated agents. These agents participate in virtual propagation to facilitate early fake news detection.
• Figure 3: Ablation study of AVOID on three datasets.
• Figure 4: The effect of different hyparameters $\lambda_{\text{rec}}$ and $\lambda_{\text{skl}}$ on three datasets.
• Figure 5: Comparative results of various LLM-based fake news detection model of accuracy and token consumption across three datasets.