CAS-SIAT-XinHai/FontHalu

该数据集涉及多模态大语言模型中的基于字体的幻觉问题。详细描述了基准构建过程，包括字体选择、模板创建、内容生成和基准组装。数据集包含VarRec、SemAna和SentAna等任务，具体说明了每个任务的图像和问题数量。该数据集用于评估各种模型在这些任务上的性能。描述全面，涵盖了方法学、数据类型和质量控制措施。

This dataset pertains to font-based hallucinations in multimodal large language models. It elaborates on the benchmark construction process, which includes font selection, template creation, content generation, and benchmark assembly. The dataset encompasses tasks such as VarRec, SemAna, and SentAna, with specific details on the number of images and questions for each task. The dataset is utilized to evaluate the performance of various models on these tasks. The description is thorough, covering the methodology, data types, and quality control measures.