The paper introduces AgriTune-R, a reproducible and auditable framework designed to adapt general-purpose large language models for specific agricultural applications. This approach addresses the domain-specific, safety-critical nature of agriculture by integrating data governance, expert evaluation, and evidence constraints to prevent unreliable advice.
- The framework uses Qwen3-8B as the recommended base model and employs LoRA/QLoRA parameter-efficient fine-tuning alongside retrieval-augmented generation.
- It establishes an evaluation protocol covering agricultural knowledge QA, pest consultation, cultivation management, and policy explanation.
- An expert-review rubric is provided that assesses factuality, safety, evidence consistency, and uncertainty expression.
- The work clearly separates protocol design from empirical conclusions to provide an executable baseline for future studies.
This framework provides a structured workflow and clear separation between design and results, offering a verifiable baseline for developing safe and accurate agricultural AI assistants.