stefanocarrera/autophagycode_D_he_train-mercury_Qwen3-8B_strategy_trust_t0.75_g2_run1_metrics
收藏资源简介:
该数据集是一个用于编程任务评估和代码分析的数据集,包含164个训练样本。每个样本代表一个编程任务,具有多个特征,包括任务ID、入口点、可执行状态、正确性标志、通过和失败的测试数量、测试运行时间(当前为null)、错误类型,以及一系列软件度量指标,如Halstead复杂度度量(词汇量、长度、体积、难度、努力程度、时间)、圈复杂度、可维护性指数、代码行数(LOC和SLOC)、注释百分比、类型标记比率(TTR)、令牌字典、香农熵、平均预测熵、最大预测熵、定义函数数量,以及入口点是否重复的布尔标志。这些特征可用于评估代码质量、复杂性和执行性能,适用于软件工程、机器学习模型训练或代码生成任务的研究。
This dataset is designed for programming task evaluation and code analysis, containing 164 training examples. Each example represents a programming task with multiple features, including task ID, entry point, executable status, correctness flag, number of tests passed and failed, test run time (currently null), error type, and a set of software metrics such as Halstead complexity measures (vocabulary, length, volume, difficulty, effort, time), cyclomatic complexity, maintainability index, lines of code (LOC and SLOC), comment percentage, type-token ratio (TTR), token dictionary, Shannon entropy, mean predictive entropy, max predictive entropy, number of functions defined, and a boolean flag for entry point repetition. These features are useful for assessing code quality, complexity, and execution performance, and can be applied in software engineering research, machine learning model training, or code generation tasks.




