QCRI/MemeLens
收藏资源简介:
MemeLens是一个大规模多语言多模态的梗图理解基准数据集,包含46个分类任务,涵盖9种语言(阿拉伯语、孟加拉语、德语、英语、西班牙语、印地语、罗马尼亚语、俄语和中文)。数据集通过LLM生成的解释和LLM-as-Judge的质量评分进行了丰富。数据集结构按语言和任务组织,包含训练、测试和验证集,每个样本包含唯一的标识符、图像路径、OCR文本、分类标签、任务描述、LLM生成的解释等字段。测试集还包括LLM-as-Judge的评分和解释。
MemeLens is a large-scale multilingual multimodal meme understanding benchmark with 46 classification tasks across 9 languages (Arabic, Bengali, German, English, Spanish, Hindi, Romanian, Russian, and Chinese). The dataset is enriched with LLM-generated explanations and LLM-as-Judge quality scores. It is organized by language and task, with splits for training, testing, and validation. Each sample includes a unique identifier, image path, OCR text, classification label, task description, LLM-generated explanation, and more. The test set also includes LLM-as-Judge scores and justifications.




