2025/06/10 by Huixuan Zhang, Xiaojun Wan, Zhang, Huixuan +1
Computer Science · Arts and Humanities · #Multimodal Machine Learning Applications #Generative Adversarial Networks and Image Synthesis #Digital Humanities and Scholarship
paper · pdf · doi:10.48550/arxiv.2506.08480
Text-to-image models often struggle to generate images that precisely match textual prompts. Prior research has extensively studied the evaluation of image-text alignment in text-to-image generation. However, existing evaluations primarily focus on agreement with human assessments, neglecting other critical properties of a trustworthy evaluation framework. In this work, we first identify two key aspects that a reliable evaluation should address. We then empirically demonstrate that current mainstream evaluation frameworks fail to fully satisfy these properties across a diverse range of metrics and models. Finally, we propose recommendations for improving image-text alignment evaluation.