AutomatedScientist/jupyter-interleaved-tool-call

该数据集是Aurora-M笔记本语料库的一部分，专门用于文本生成任务。它处理来自Jupyter笔记本的JSONL记录和围栏代码块，将Markdown围栏代码块转换为显式的工具调用/工具响应标记，并使用Python的AST对可解析性进行评分。数据集包含用于分析、转换、包分析和评分的脚本，支持并行处理，并提供make目标来运行整个流程。README还提供了设置说明、数据布局以及用于检查结果的查看器。

This dataset is part of the Aurora-M notebook corpus and is designed for text-generation tasks. It processes JSONL records and fenced code blocks from Jupyter notebooks, transforming markdown fenced code blocks into explicit tool-call/tool-response markers and scoring their parseability using Pythons AST. The dataset includes scripts for analysis, transformation, package analysis, and rubric scoring, with support for parallelism and make targets for running the pipeline. The README also provides setup instructions, data layout, and a viewer for inspecting the results.