North-ML1/CodeBench-30
收藏资源简介:
CodeBench-30是一个包含30个编程问题的基准数据集,旨在评估AI模型在不同语言、领域和难度级别上的编码能力。数据集包含10个简单、10个中等和10个困难的问题,支持Python、JavaScript和SQL语言。每个问题包含唯一标识符、难度级别、语言、类别标签、问题描述、示例输入输出、测试用例和参考解决方案。评估方法是通过比较模型生成的代码输出与预期输出。数据集还提供了按难度、语言和问题类别的详细分类,以及加载和使用数据集的示例代码。
CodeBench-30 is a benchmark dataset of 30 coding problems designed to evaluate AI model coding ability across languages, domains, and difficulty levels. The dataset includes 10 easy, 10 medium, and 10 hard problems, supporting Python, JavaScript, and SQL. Each problem contains a unique identifier, difficulty level, language, category tag, problem statement, example input/output pairs, test cases, and a reference solution. Evaluation is performed by comparing the models generated code output to the expected output. The dataset also provides detailed breakdowns by difficulty, language, and problem category, along with example code for loading and using the dataset.




