VLM Diagnosis of Visual Persuasiveness and Recall-Oriented Bias

Evaluation of Vision-Language Models reveals a recall-oriented bias where models over-predict images as persuasive compared to human raters. The study introduces Visual Persuasive Factors (VPFs), a cognitive psychology-informed taxonomy to better quantify how visual cues influence persuasive judgment.