THEMISTO
收藏资源简介:
THEMISTO数据集由JetBrains Research创建,包含Jupyter笔记本开发轨迹,用于评估大型语言模型在预测代码输出和代码生成任务中利用运行时信息的能力。该数据集由29个笔记本的超过9000次单元格执行组成,不仅包括环境和笔记本的最终版本,还包括笔记本中的中间和调试步骤,这些是模型应支持开发者的关键阶段。数据集旨在解决代码语言模型中融入运行时信息的挑战,推动该领域的进一步研究。
The THEMISTO dataset was created by JetBrains Research. It comprises Jupyter notebook development trajectories, designed to evaluate the ability of large language models (LLMs) to leverage runtime information for code output prediction and code generation tasks. This dataset includes over 9,000 cell executions across 29 notebooks, covering not only the final versions of the execution environments and notebooks, but also the intermediate and debugging steps within the notebooks—critical stages where models should support developers. The dataset aims to address the challenges of integrating runtime information into code language models, and advance further research in this field.

- 1Themisto: Jupyter-Based Runtime BenchmarkJetBrains Research · 2025年



