MBZUAI/DuwatBench

DuwatBench是一个全面的基准数据集，用于评估语言模型在阿拉伯书法识别上的表现。阿拉伯书法代表了阿拉伯语言最丰富的视觉传统之一，将语言意义与艺术形式相结合。DuwatBench填补了现代AI系统处理风格化阿拉伯文本评估的空白。数据集包含1,272个精选样本，涵盖6种古典和现代书法风格，超过9.5k个单词实例和约1,475个独特单词，涉及宗教和文化领域。此外，数据集还提供了边界框注释、完整文本转录、风格和主题标签，以及复杂的艺术背景，以保持现实世界的视觉复杂性。

DuwatBench is a comprehensive benchmark for evaluating LMMs on Arabic calligraphy recognition. Arabic calligraphy represents one of the richest visual traditions of the Arabic language, blending linguistic meaning with artistic form. DuwatBench addresses the gap in evaluating how well modern AI systems can process stylized Arabic text. The dataset contains 1,272 curated samples spanning 6 classical and modern calligraphic styles, over 9.5k word instances with approximately 1,475 unique words spanning religious and cultural domains. It also includes bounding box annotations, full text transcriptions with style and theme labels, and complex artistic backgrounds preserving real-world visual complexity.