What is Sinong
Sinong (Sinong Large Language Model) is a vertical large model for the agricultural field launched by Nanjing Agricultural University. Leveraging the disciplinary strengths of Nanjing Agricultural University, the model has collected over 4 billion tokens of data covering multiple agricultural disciplines, including books, papers, policies, and more. The model is trained through methods such as synthetic data construction, instruction fine-tuning, and reinforcement learning, and possesses strong agricultural knowledge understanding capabilities. The Sinong model has open-sourced two versions, 8B and 32B, and supports multi-agent retrieval.
Main Features of Sinong
-
Agricultural Knowledge Q&A: The model can provide accurate and detailed answers to agriculture-related questions, covering multiple fields such as planting, breeding, and agricultural economics.
-
Multi-Agent Retrieval Augmentation: By optimizing the retrieval framework, it quickly extracts relevant information from massive agricultural literature, improving the efficiency of knowledge acquisition.
-
Chain-of-Thought and Context Understanding: It supports Chain-of-Thought Question Answering (COT-QA) and In-Context Question Answering (Incontext-QA), and can perform in-depth reasoning and analysis in combination with context.
-
Instruction Understanding and Execution in the Agricultural Domain: Through instruction fine-tuning, it accurately understands and executes complex instructions in the agricultural domain, assisting agricultural decision-making.
-
Data-Driven Decision Support: Combined with agricultural big data, it provides data support and decision-making suggestions for agricultural production, management, investment, and more.
Project Address of Sinong
- GitHub Repository：https://github.com/njauzzx/Sinong
Core Advantages of Sinong
- Domain Focus: Sinong focuses on the agricultural domain, with deep optimization for subdisciplines such as animal science, plant protection, and agricultural economics. It can accurately understand and handle agriculture-related issues and provide professional and accurate solutions.
- Large-Scale High-Quality Data: Relying on the disciplinary strengths of Nanjing Agricultural University, Sinong has collected more than 4 billion tokens of agricultural domain data, covering multiple types such as books, papers, policies, and patents. The data is rich and has undergone strict screening and organization, ensuring a solid foundation for model training.
- Multimodal Fusion Technology: By combining vision models with large language models, Sinong can process multiple data forms such as images and text, realize the fusion of multimodal information, and enhance the ability to understand and analyze complex agricultural scenarios.
- Instruction Fine-Tuning and Reinforcement Learning: Through synthetic data construction, instruction fine-tuning, and reinforcement learning, Sinong performs excellently in understanding and executing instructions in the agricultural domain. It can quickly adapt to diverse tasks in the agricultural domain, such as question answering, reasoning, and decision support.
- Multi-Agent Retrieval Augmentation: Sinong introduces a multi-agent retrieval augmentation framework, optimizes knowledge base construction and retrieval efficiency, and can quickly extract relevant information from massive agricultural literature, improving the efficiency of knowledge acquisition and application.
Application Scenarios of Sinong
- Agricultural Production and Planting: It provides farmers with planting plans, pest and disease control, and agricultural technology consultation, supporting scientific planting and efficient management.
- Livestock and Poultry Breeding: It assists breeders in breeding management, animal health diagnosis, and breeding environment optimization, improving breeding efficiency.
- Agricultural Economics and Market Analysis: It analyzes market trends, interprets agricultural policies, and provides suggestions for farmers and enterprises, supporting agricultural economic decision-making.
- Agricultural Research and Education: The model supports researchers' literature retrieval and experimental design, generates educational resources, and promotes the development of agricultural research and education.
- Smart Agriculture and Digital Transformation: Combined with the Internet of Things to enable intelligent monitoring and early warning, analyze agricultural big data, and optimize agricultural supply chain management.