TempestTeam/dataset-eval

dataset-eval是一个多语言、多领域的评估数据集，包含来自英语、法语、西班牙语以及Python、Java、JavaScript、C/C++等编程语言的高质量文本或代码例子。这些例子用于在语言模型训练期间进行性能评估，包括性能跟踪、跨语言或领域的泛化诊断以及实现早期停止机制。数据集不适用于直接训练，因为其规模有限且目的是作为过滤后的评估样本。

The `dataset-eval` is a multilingual and multi-domain evaluation dataset containing high-quality text or code examples from English, French, Spanish, as well as Python, Java, JavaScript, C/C++ programming languages. These examples are used for performance evaluation during language model training, including performance tracking, generalization diagnostics across languages or domains, and implementing early stopping mechanisms. The dataset is not intended for direct training due to its limited size and purpose as a filtered evaluation sample.