Fine Tuned Language Model

AbstractUnderstanding image advertisements is a challenging task, often requiring non-literal interpretation. We argue that standard image-based predictions are insufficient for symbolism prediction. Following the intuition that texts and images are …...

AbstractIn this paper, we present our submission for subtask A of the Common Sense Validation and Explanation (ComVE) shared task. We examine the ability of large-scale pre-trained language models to …...