# AgroLLM: Connecting Farmers and Agricultural Practices through Large Language Models for Enhanced Knowledge Transfer and Practical Application

**Authors:** Dinesh Jackson Samuel, Inna Skarga-Bandurova, David Sikolia, Muhammad Awais  
**arXiv:** [2503.04788](https://arxiv.org/abs/2503.04788)  
**Date:** 2025-02-28  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2503.04788v1/

## Key points

- AgroLLM is an AI-powered chatbot that uses Large Language Models (LLMs) and a Retrieval-Augmented Generation (RAG) framework to enhance knowledge-sharing in agriculture.
- It taps into an open-source agricultural database for accurate and contextually relevant responses, minimizing the risk of incorrect information retrieval.
- Utilizes the FAISS vector database for efficient similarity searches, ensuring swift access to agricultural knowledge.
- A comparative study evaluated AgroLLM's performance across key agricultural domains using advanced models like ChatGPT-4o Mini with RAG achieving 93% accuracy rate.
- Continuous feedback mechanisms help enhance response quality over time, making AgroLLM a valuable educational tool for farmers, researchers, and professionals in the agriculture industry.
- The project aims to develop AgroLLM as a centralized knowledge base LLM tailored for agriculture, showcasing a structured approach to integrating LLMs into agricultural applications.
- AgroLLM bridges the gap between theoretical knowledge and practical applicability for farmers, setting a new standard for delivering accurate and contextually relevant agricultural information.

## Summary

AgroLLM is an innovative AI-powered chatbot that leverages Large Language Models (LLMs) and a Retrieval-Augmented Generation (RAG) framework to enhance knowledge-sharing and education in the field of agriculture. By tapping into a comprehensive open-source agricultural database, AgroLLM provides accurate and contextually relevant responses while minimizing the risk of incorrect information retrieval. The system utilizes the FAISS vector database for efficient similarity searches, ensuring swift access to a wealth of agricultural knowledge. To evaluate AgroLLM's performance across key agricultural domains such as Agriculture and Life Sciences, Agricultural Management, Agriculture and Forestry, and Agriculture Business, a comparative study was conducted using three advanced models: Gemini 1.5 Flash, ChatGPT-4o Mini, and Mistral-7B-Instruct-v0.2. Key evaluation metrics included embedding quality, search efficiency, and response relevance. Results showed that ChatGPT-4o Mini with RAG achieved an impressive accuracy rate of 93%. One standout feature of AgroLLM is its continuous feedback mechanisms which help enhance response quality over time. This makes AgroLLM not just a valuable educational tool for farmers, researchers, and professionals in the agriculture industry but also a benchmark for AI-driven solutions aimed at promoting informed decision-making and improving agricultural practices. The project aims to develop AgroLLM as a centralized knowledge base LLM specifically tailored for agriculture. By showcasing a structured approach to integrating LLMs into agricultural applications - from data collection and preprocessing to embedding generation, retrieval using vector databases, and retrieval-augmented generation - this work serves as a proof of concept that can inspire further research and development in this domain. Through its unique ability to bridge the gap between theoretical knowledge and practical applicability for farmers, AgroLLM stands out as an invaluable resource in the realm of agriculture-specific AI technologies. With its foundation built on cutting-edge AI technologies like generative AI for text extraction and FAISS format for efficient similarity searches, AgroLLM sets a new standard for delivering accurate and contextually relevant agricultural information. The overall methodology outlined in this project encompasses three main components: data collection and processing, development of the AgroLLM framework, and evaluation/selection of LLM models best suited for generating responses related to farming and agriculture. This comprehensive approach ensures that AgroLLM remains at the forefront of AI-driven solutions in agriculture by continuously refining its capabilities through rigorous evaluation processes.