AgriPestDatabase-v1.0 Released to Train Compact AI Models for Farm-Level Pest Management
Researchers have released AgriPestDatabase-v1.0, a structured insect dataset designed to train lightweight language models for agricultural pest management, according to a paper accepted at the Artificial Super Intelligence Conference 2026.
According to arxiv.org, the work addresses two key challenges: generating structured insect information and adapting compact LLMs (≤7B parameters) for deployment on edge devices in rural areas with limited internet connectivity. The researchers collected textual data by reviewing pest databases and published manuscripts on nine selected pest species, which were then validated by a domain expert and converted into question-answer pairs for model training.
Using LoRA-based fine-tuning, the team evaluated multiple lightweight models. According to the paper, Mistral 7B achieved an 88.9% pass rate on domain-specific tasks, “substantially outperforming Qwen 2.5 7B (63.9%), and LLaMA 3.1 8B (58.7%).” The research found that Mistral demonstrated higher semantic alignment (embedding similarity: 0.865) despite lower lexical overlap (BLEU: 0.097), indicating that “semantic understanding and robust reasoning are more predictive of task success than surface-level conformity in specialized domains,” according to the paper.
The work aims to provide practical decision support tools for farmers through compact, deployable systems that can function in field-level conditions without continuous internet access.