Drifting Away from Truth: GenAI-Driven News Diversity Challenges LVLM-Based Misinformation Detection

TL;DR

This paper reveals how GenAI-driven news diversity causes multi-level drift that significantly impairs the robustness of LVLM-based misinformation detection systems, highlighting critical vulnerabilities and the need for more resilient methods.

Contribution

The paper introduces DriftBench, a large-scale benchmark for evaluating LVLM-based misinformation detectors under diverse and adversarial GenAI-generated content, and systematically analyzes their vulnerabilities.

Findings

LVLM-based detectors suffer an average F1 drop of 14.8% under drift.

Current systems are highly susceptible to adversarial evidence contamination.

Reasoning consistency deteriorates with increased content diversity.

Abstract

The proliferation of multimodal misinformation poses growing threats to public discourse and societal trust. While Large Vision-Language Models (LVLMs) have enabled recent progress in multimodal misinformation detection (MMD), the rise of generative AI (GenAI) tools introduces a new challenge: GenAI-driven news diversity, characterized by highly varied and complex content. We show that this diversity induces multi-level drift, comprising (1) model-level misperception drift, where stylistic variations disrupt a model's internal reasoning, and (2) evidence-level drift, where expression diversity degrades the quality or relevance of retrieved external evidence. These drifts significantly degrade the robustness of current LVLM-based MMD systems. To systematically study this problem, we introduce DriftBench, a large-scale benchmark comprising 16,000 news instances across six categories of diversification. We design three evaluation tasks: (1) robustness of truth verification under multi-level drift; (2) susceptibility to adversarial evidence contamination generated by GenAI; and (3) analysis of reasoning consistency across diverse inputs. Experiments with six state-of-the-art LVLM-based detectors show substantial performance drops (average F1 -14.8%) and increasingly unstable reasoning traces, with even more severe failures under adversarial evidence injection. Our findings uncover fundamental vulnerabilities in existing MMD systems and suggest an urgent need for more resilient approaches in the GenAI era.