CAMEL-Bench
收藏资源简介:
CAMEL-Bench是由穆罕默德·本·扎耶德人工智能大学开发的全面阿拉伯语大型多模态模型评估基准,旨在代表超过4亿阿拉伯语使用者。该数据集包含29,036个问题,涵盖八个不同领域和38个子领域,包括多图像理解、复杂视觉感知、手写文档理解、视频理解、医学影像、植物疾病和遥感土地使用理解等。数据集的创建过程包括从现有英语基准翻译和手动验证,确保高质量。CAMEL-Bench主要应用于评估和改进阿拉伯语多模态模型的性能,特别是在视觉推理和理解任务中。
CAMEL-Bench is a comprehensive Arabic large multimodal model evaluation benchmark developed by Mohamed bin Zayed University of Artificial Intelligence, designed to cater to the over 400 million Arabic speakers worldwide. The dataset contains 29,036 questions spanning 8 distinct domains and 38 sub-domains, including multi-image understanding, complex visual perception, handwritten document understanding, video understanding, medical imaging, plant disease recognition, remote sensing land use interpretation, and more. The dataset was constructed through translation of existing English benchmarks and manual validation, ensuring high data quality. CAMEL-Bench is primarily applied to evaluate and enhance the performance of Arabic multimodal models, particularly in visual reasoning and understanding tasks.

- 1CAMEL-Bench: A Comprehensive Arabic LMM Benchmark穆罕默德·本·扎耶德人工智能大学 · 2024年



