Synthetic Agricultural Datasets for Computational Stress-Testing (CROP-LENS Framework)

These synthetic datasets were created  to test the algorithmic scalability and high-volume robustness (up to 3,500+ rows) of the CROP-LENS framework. 

These datasets were created utilizing a Large Language Model initialized with actual agricultural data to ensure believable minimum and maximum limits for characteristics such as Temperature and Soil pH. Nonetheless, generative LLMs do not accurately represent genuine environmental multi-collinearity. For instance, these synthetic datasets include fabricated statistical associations (e.g., a created >0.80 correlation between humidity and rainfall, and inverted chemical correlations between Potassium and Phosphorous).

Consequently, this information should not be utilized for any biological, ecological, or agronomic assessments. It is offered solely for the purposes of computational reproducibility and rigorous software testing.

本批合成数据集专为测试CROP-LENS框架的算法可扩展性与高容量鲁棒性（最高支持3500+行数据）而构建。
本数据集通过基于真实农业数据初始化的大语言模型（Large Language Model）生成，以确保温度、土壤pH值等特征的取值范围具备合理性。但生成式大语言模型无法准确还原真实环境中的变量多重共线性特征。例如，本合成数据集包含人为构造的统计关联：如湿度与降雨量之间被设定为大于0.80的相关性，以及钾元素与磷元素之间的化学关联呈现反向关系。
因此，本数据集不得用于任何生物学、生态学或农艺学相关评估，仅用于计算可重复性验证与严格的软件测试工作。