NLEBench

NLEBench是专为评估挪威语生成语言模型能力而设计的综合基准数据集，由挪威科技大学计算机科学系创建。该数据集包含5000个样本，涵盖从新闻叙事、摘要、开放领域对话到自然语言理解等多个真实世界NLP任务。数据集特色在于包含两个高质量人工标注数据集：一个覆盖挪威传统文化、习语、俚语和特殊表达的指令数据集，以及一个基于文档的多标签数据集，用于主题分类、问答和摘要。NLEBench旨在通过这些多样化的任务，评估和揭示主流语言模型在低资源语言如挪威语中的独特特性和能力，推动针对此类语言的更先进语言模型研究。

NLEBench is a comprehensive benchmark dataset specifically designed to evaluate the capabilities of Norwegian language generation models, created by the Department of Computer Science at the Norwegian University of Science and Technology. This dataset contains 5,000 samples, covering a wide range of real-world NLP tasks including news narration, text summarization, open-domain dialogue, and natural language understanding. What sets NLEBench apart are two high-quality manually annotated datasets: one instruction dataset covering Norwegian traditional culture, idioms, slang and specialized expressions, and another document-based multi-label dataset for topic classification, question answering and summarization. NLEBench aims to assess and reveal the unique characteristics and capabilities of mainstream language models in low-resource languages such as Norwegian, and promote the development of more advanced language models targeting such languages.