Large-Scale Dataset of Porous Carbon Materials for Supercapacitors Extracted via Large Language Models

This dataset comprises a large-scale collection of experimental data on porous carbon supercapacitors, extracted from approximately 1,000 high-content research articles using an automated Large Language Model (LLM) mining framework.
The database covers over 5,000 distinct carbon samples and includes more than 10,000 specific capacitance data points under varying testing conditions. It offers a holistic view of the material properties, integrating synthesis parameters (preparation crafts), microstructural characteristics (pore size distribution, surface area), and surface elemental composition. Furthermore, it details key electrochemical performance metrics, including Specific Capacitance, Energy Density, Power Density, and Equivalent Series Resistance (ESR).
This dataset serves as a valuable resource for data-driven materials science, enabling quantitative analysis of structure-performance correlations and the inverse design of high-performance energy storage materials.
Note: For further modeling use, detailed data cleaning may be required for specific domains

The ml_use_data folder contains two datasets: the number_metric_dataset_cleaned.csv (9,962 entries), which serves as the cleaned feature pool, and the train_test_dataset.csv (284 entries), a subset derived by filtering out non-null features for model training and testing.