PandasPlotBench
收藏资源简介:
PandasPlotBench是由JetBrains Research开发的人工策划数据集,旨在评估语言模型在数据可视化探索中的辅助效果。该数据集包含175个独特的任务,专注于根据自然语言指令生成Pandas DataFrame的绘图代码。数据集通过从Matplotlib图库中提取数据并进行处理生成,确保任务和数据文件是合成生成的,以避免数据泄露。PandasPlotBench的应用领域主要集中在数据分析和可视化,旨在解决复杂任务中生成可执行绘图代码的挑战。
PandasPlotBench is a manually curated dataset developed by JetBrains Research, designed to evaluate the auxiliary capabilities of language models in data visualization exploration. This dataset comprises 175 unique tasks focused on generating plotting code for Pandas DataFrames based on natural language instructions. The dataset is constructed by extracting and processing data from the Matplotlib gallery, ensuring that both the tasks and their associated data files are synthetically generated to prevent data leakage. Primarily targeted at the fields of data analysis and visualization, PandasPlotBench aims to address the challenge of generating executable plotting code for complex tasks.

- 1Drawing Pandas: A Benchmark for LLMs in Generating Plotting CodeJetBrains Research · 2024年



