FusionBench

FusionBench是由武汉大学等机构创建的综合性深度模型融合基准数据集，包含26个不同的任务，涵盖开放词汇图像分类、文本分类和文本到文本生成等多个领域。数据集通过精心设计的模型融合技术，如模型集成、模型合并和模型混合，来评估和提升模型的性能。FusionBench旨在通过标准化评估，解决深度模型融合技术评估不一致和不足的问题，推动深度学习领域的发展。

FusionBench is a comprehensive deep model fusion benchmark dataset created by Wuhan University and other institutions. It includes 26 distinct tasks spanning multiple domains such as open-vocabulary image classification, text classification, and text-to-text generation. The dataset adopts carefully designed model fusion techniques including model ensembling, model merging and model mixing to evaluate and improve model performance. FusionBench aims to address the inconsistencies and inadequacies in the evaluation of deep model fusion technologies via standardized assessment, so as to promote the development of the deep learning field.