AI system detects and explains mismatches between images and claims
MIC: Explaining Image-Claim Inconsistencies in AI-Generated Multimodal Misinformation
Computation and Language
Summary
AI-generated images are often paired with false or misleading claims, making it hard to tell what is true. Existing tools mostly detect if an image is AI-made but don’t check if the image actually matches the claim it’s paired with. The authors created a system called MIC that helps fact-checkers find and explain when images and claims don’t match using real-world knowledge. They also made a large test set showing pairs of real and AI-altered images with different claims to evaluate their system’s ability to detect these mismatches.
What this means in practice
- •For social media content moderators: Detect posts where AI-generated images do not match their claims and provide explanations to improve content review workflows.
- •For news verification teams: Assist human fact-checkers by highlighting inconsistencies between images and accompanying claims that may indicate misinformation.
Authors
Ruihong Zeng, Jonathan Tonglet, Preslav Nakov, Iryna Gurevych
Abstract
Claims paired with AI-generated images are a rapidly growing form of misinformation. Existing automated fact-checking (AFC) methods mainly treat this as a provenance problem, detecting low-level synthesis artifacts to decide whether an image is AI-generated. However, such methods do not verify what human fact-checkers often check: whether an image's content is consistent with the context implied by its accompanying claim. To address this gap, we introduce MIC (Multimodal Inconsistency Checking), an AFC framework that assists human fact-checkers by detecting AI-generated multimodal misinformation and explaining inconsistencies using world knowledge. MIC first uses supervised fine-tuning (SFT) for task adaptation and then applies Group Relative Policy Optimization (GRPO) to directly optimize component-level verifiable rewards for verdict prediction, inconsistency type classification, visual evidence description, and world-knowledge explanation. We further introduce MIC-Bench, a benchmark comprising 8,812 image-claim instances derived from 4,406 claims, where each claim is paired with an authentic image and an AI-generated counterpart that introduces a controlled contextual inconsistency. Compared with SFT alone, GRPO further improves Macro-F1 by 4.67 and 4.11 points in the in-distribution and out-of-distribution settings, respectively, while also improving the semantic similarity of visual evidence descriptions and world-knowledge explanations to reference annotations. Our code and data are available at https://github.com/UKPLab/arxiv2026-mic.