mopsa-dataset-debug
收藏资源简介:
Mopsa数据集是一个统一的、包含静态分析跟踪、具体执行跟踪以及生成的自然语言跟踪的数据集。数据集的核心是源代码及其对应的多层次语义表示。每个数据样本包含原始源代码、经过清理的Mopsa静态分析跟踪、用于对齐跟踪的代码片段,以及将静态注释与这些代码片段对齐后的结果。此外,数据集还提供了具体的执行跟踪(包括原始和投影后的状态),以及将这些执行状态与代码片段对齐的结果。数据集记录了样本的来源数据集名称、唯一标识符、最终输出的变量名及其预期的输出映射。为了增强可解释性,部分样本还包含了静态跟踪和执行跟踪的自然语言重写版本,以及分别与静态跟踪片段和执行跟踪片段对齐的、交错排列的自然语言解释列表。该数据集适用于程序理解、代码语义分析、执行行为预测以及生成代码或程序状态的文本解释等研究任务。
The Mopsa dataset is a unified dataset containing static analysis traces, concrete execution traces, and generated natural language traces. The core of the dataset is source code and its corresponding multi-level semantic representations. Each data sample includes original source code, cleaned Mopsa static analysis traces, code snippets for aligning traces, and the results of aligning static annotations with these code snippets. Additionally, the dataset provides concrete execution traces (including original and projected states) and the results of aligning these execution states with code snippets. The dataset records the source dataset name, unique identifier, final output variable names, and their expected output mappings for each sample. To enhance interpretability, some samples also include natural language rewrites of static and execution traces, as well as interleaved lists of natural language explanations aligned with static and execution trace segments. This dataset is suitable for research tasks such as program understanding, code semantic analysis, execution behavior prediction, and generating textual explanations of code or program states.




