A parameter-efficient three-branch architecture for multimodal misinformation detection with limited annotations

With the rise of easily accessible generative tools for creating and manipulating multimedia content, the threat of realistic synthetic alterations to digital media, often involving manipulations across multiple modalities simultaneously, has grown exponentially. Recently, such techniques have been