lmms-lab/EMMA
收藏Hugging Face2025-07-25 更新2025-08-09 收录
下载链接：
https://hf-mirror.com/datasets/lmms-lab/EMMA
下载链接
链接失效反馈官方服务：
资源简介：
EMMA（增强多模态推理）是一个针对数学、物理、化学和编程领域的有机多模态推理基准测试。EMMA任务需要高级的跨模态推理能力，不能仅通过单独考虑每个模态来解决，为机器学习语言模型提供了增强的推理能力测试套件。数据集由2788个问题组成，涵盖了四个领域，每个问题都根据其测量的特定技能进行了细粒度标注。
EMMA (Enhanced MultiModal reAsoning) is a benchmark targeting organic multimodal reasoning across mathematics, physics, chemistry, and coding. EMMA tasks require advanced cross-modal reasoning capabilities that cannot be solved by considering each modality separately, offering an enhanced test suite for MLLMs reasoning abilities. The dataset consists of 2,788 problems covering four domains, with each problem finely labeled based on the specific skills it measures.
提供机构：
lmms-lab
