EthanYan12581/RobustStress

RobustStress是一个由武汉大学策划的英文数据集，可能由youcheng Yan共享，采用apache-2.0许可证。该数据集旨在评估模型的鲁棒性，包含多种结构分类，如跨领域跨模型、跨领域模型特定、领域特定跨模型和领域特定模型特定。它还涉及攻击方法，包括基于大型语言模型的释义攻击和同义词替换。数据集规模在10亿到100亿个元素之间，专注于自然语言处理任务中的压力测试和鲁棒性分析。RobustStress is an English dataset curated by Wuhan University, possibly shared by youcheng Yan, under the apache-2.0 license. It is designed to evaluate model robustness, featuring structural categories such as cross_domains_cross_models, cross_domains_model_specific, domain_specific_cross_models, and domain_specific_model_specific. The dataset includes attack methods like LLM-based paraphrasing attack and synonym replacement. With a size between 1B and 10B elements, it focuses on stress testing and robustness analysis in NLP tasks.