RoCOCO：鲁棒性基准 MS-COCO 以测试图像文本匹配模型的鲁棒性（英文版）

英文标题：RoCOCO: Robust Benchmark MS-COCO to Stress-test Robustness of Image-Text Matching Models中文摘要：本文提出了一个新的评估基准来测试 ITM 模型的稳健性，加入了各种欺骗性图像和标题来检测模型。结果发现，添加这些图像和标题会降低 SOTA 模型的性能，预计这些结果能提高模型的鲁棒性和引入更多的评估方法。英文摘要：Recently, large-scale vision-language pre-training models and visual semanticembedding methods have significantly improved image-text matching (ITM)accuracy on MS COCO 5K test set. However, it is unclear how robust thesestate-of-the-art (SOTA) models are when using them in the wild. In this pa