Table 2.

Comparative analysis: crop-specific LLMs, general foundation models, fine-tuning, RAG, and KG-grounded architectures.

S. no.	Approach	Main advantage	Main limitation	Scalability	Computation cost	Best use
1	Crop-specific LLM	Strong domain accuracy	Narrow generalization	Low–medium	High	Rice biology, pest/disease advisory
2	General foundation model	Broad reasoning and flexibility	Weak crop-specific grounding	High	Low–medium	General advisory, multilingual Q&A
3	Fine-tuning	Internalizes crop knowledge	Overfitting and retraining burden	Medium	High	Stable domain tasks
4	RAG	Easy knowledge updating	Depends on retrieval quality	High	Medium	Dynamic advisory and extension
5	KG-grounded LLM	Traceable, explainable reasoning	KG construction and maintenance burden	Medium	Medium–high	High-stakes agronomic decisions