MLLM-CL/MLLM-CL-ReplayData

MLLM-CL是一个用于多模态大型语言模型持续学习的基准数据集。它包括两个主要组件：领域持续学习和能力持续学习。领域持续学习涉及在五个主流和常见领域（遥感、医疗、科学、自动驾驶和金融）中持续添加领域知识。能力持续学习涉及学习四个基本能力：OCR、数学和逻辑、视觉感知和GUI代理。该数据集支持英语，并使用Apache-2.0许可证。数据集大小小于1K，任务类别为图像-文本到文本。

MLLM-CL is a benchmark dataset for continual learning in multimodal large language models. It includes two main components: domain continual learning and ability continual learning. Domain continual learning involves continuously adding domain knowledge across five mainstream domains: remote sensing, medical, science, autonomous driving, and finance. Ability continual learning involves learning four fundamental abilities: OCR, math & logic, visual perception, and GUI agent. The dataset supports English and is licensed under Apache-2.0. The dataset size is less than 1K, and the task category is image-text-to-text.