FineCops-Ref
收藏资源简介:
FineCops-Ref是一个专为多模态大型语言模型设计的细粒度指代表达式理解数据集,由电子科技大学的研究团队创建。该数据集包含正样本和负样本,正样本由图像和文本表达式组成,负样本则是未配对的图像和文本表达式,用于评估模型在目标物体不存在的场景下的拒绝能力。数据集根据细粒度推理的复杂度分为三个难度级别,通过引入负样本,全面评估模型的视觉定位能力。
FineCops-Ref is a fine-grained referring expression comprehension dataset specifically designed for multimodal large language models, developed by the research team from the University of Electronic Science and Technology of China. This dataset comprises positive and negative samples. Positive samples are image-text pairs, while negative samples refer to unpaired image and text expressions, which are utilized to assess the model's ability to reject scenarios where the target object is absent. The dataset is categorized into three difficulty levels based on the complexity of fine-grained reasoning, and comprehensively evaluates the visual grounding capability of models by introducing negative samples.




