Thread Content
Recently, the \"General High-Quality Dataset for the Petrochemical Industry\" developed by Sinopec successfully passed the evaluation conducted by the Electronic Industry Standardization Institute of the Ministry of Industry and Information Technology, becoming the first high-quality industry dataset in China to pass such a standard evaluation. This evaluation is based on the **standard \"Specifications for Quality Assessment of High-Quality Datasets\" (TC609-5-2025-04)\). It employs a assessment method that combines data and models, covering 17 indicators across 3 dimensions: dataset documentation, data quality, and model application, in order to conduct a comprehensive multi-dimensional quality assessment of the datasets. “The \"Universal High-Quality Dataset for the Petrochemical Industry\" includes industry-specific terminology, formulas, molecular formulas, etc. It scores over 90 points in terms of format standardization, content consistency, and data cleanliness, demonstrating industry-leading data quality and fully verifying its reliability in supporting the advanced applications of artificial intelligence. This dataset provides strong support for the training of Sinopec Great Wall large models, as well as for the training of specialized large models used for molecular understanding, molecular generation, and auditing. The high-quality dataset and related toolkits developed in-house facilitate data processing for nearly 900 users across 86 organizations within the system. At the same time, its standardized construction method also provides a reference model for the petrochemical industry.