China has unveiled its first open-source large language model dedicated to crop protection, marking an important step in the application of artificial intelligence for ensuring security and improving productivity in agriculture. As reported in Science and Technology Daily, the model, called Green Shield, was developed by Nanjing Agricultural University in collaboration with the National Key Laboratory of Agricultural Biosecurity and more than 30 industry partners.
The launch comes as China faces recurring pest outbreaks and growing issues related to pesticide resistance, which are increasingly constraining agricultural yields and sustainability. Project leader Dong Shameng, who also serves as deputy dean of the university's Plant Protection Institute, noted that farmers urgently need accurate and professional localized guidance. He pointed out that general-purpose large language models often cannot provide reliable answers in specialized agricultural contexts and may even give inconsistent or potentially dangerous pesticide recommendations.
To address these shortcomings, the research team created a highly specialized dataset consisting of more than 2.5 billion tokens drawn from academic literature, patents, national standards, and field reports. The dataset covers major crop categories such as rice, wheat, soybeans, vegetables, and fruit trees, and also integrates pest monitoring data, eco-friendly control measures, and pesticide registration information.
According to Professor Wang Dongbo of the university's Information Management Institute, the model is capable of accurately identifying crop types, growth stages, and disease symptoms before generating comprehensive treatment strategies. Through targeted training, the system achieves strong convergence and demonstrates high accuracy in pest recognition, enabling more reliable decision-making in field conditions.
A key safety feature of Green Shield is its automatic compliance mechanism. Before issuing any recommendations, the model cross-references inputs with the national pesticide registration database to ensure that proposed chemicals meet regulatory requirements, including approved usage, crop compatibility, and dosage restrictions. Any non-compliant outputs are automatically blocked and corrected, reducing the risk of pesticide misuse at the outset.
Wang Yuanchao, vice president of Nanjing Agricultural University, stated that the institution will continue to refine the model through field trials and iterative improvements. The goal is to create an intelligent agricultural tool that is accessible, practical, and effective for farmers, ultimately supporting the digital transformation of modern agriculture across the entire production chain.