Fine-Tuning General-Purpose Large Language Models for Agricultural Applications:A Reproducible Framework and Evaluation Protocol Based on Qwen3-8B

2026-06-27

AI Summary

Agricultural question answering is characterized by high domain specificity, regional dependency, time sensitivity, and safety-critical implications, rendering general-purpose large language models prone to generating unreliable or even hazardous advice. This work proposes AgriTune-R, a framework built upon Qwen3-8B that integrates agricultural data curation, instruction construction, efficient LoRA/QLoRA fine-tuning, and retrieval-augmented generation (RAG) through a structured adaptation pipeline. It further incorporates expert-driven safety controls and uncertainty expression mechanisms. Key contributions include a reproducible workflow for adapting large models to agriculture, a multidimensional evaluation protocol jointly assessing factual accuracy and safety, and an expert-reviewed scoring rubric. Experimental results demonstrate that AgriTune-R significantly enhances the factual correctness and safety of model outputs across tasks such as crop disease diagnosis, pesticide application, cultivation management, and agricultural policy interpretation.

Abstract

General-purpose large language models (LLMs) have demonstrated strong abilities in opendomain question answering, information extraction, and text generation. Agricultural applications, however, are domain-specific, region-dependent, time-sensitive, and safety-critical. Without data governance, expert evaluation, and evidence constraints, an agricultural assistant mayproduce unreliable advice on crop diseases, pesticide use, fertilization, or policy interpretation.To avoid presenting unverified simulated numbers as real experimental findings, this paper doesnot report any model-performance claims that have not been produced by an actual training runand expert evaluation. Instead, we propose AgriTune-R, a reproducible and auditable frameworkfor adapting general-purpose LLMs to agricultural tasks. The framework selects the publiclyverifiable Qwen3-8B model as the recommended base model and integrates agricultural datagovernance, instruction construction, LoRA/QLoRA parameter-efficient fine-tuning, retrievalaugmented generation, expert evaluation, and safety control for high-risk questions. The contributions are: (1) a structured workflow for agricultural LLM adaptation; (2) an evaluationprotocol for agricultural knowledge QA, pest and disease consultation, cultivation management,and policy explanation; (3) an expert-review rubric combining factuality, safety, evidence consistency, and uncertainty expression; and (4) a clear separation between protocol design andempirical conclusions, providing an executable baseline for future empirical studies.

Problem

• agricultural applications
• large language models
• domain-specific
• safety-critical
• data governance

Innovation

• AgriTune-R
• parameter-efficient fine-tuning
• retrieval-augmented generation
• expert evaluation
• data governance