Graph-Augmented Reasoning with Large Language Models for Tobacco Pest and Disease Management

2026-02-02

AI Summary

This study addresses the challenges of unstructured knowledge and hallucination in large language models when applied to tobacco pest and disease management. To mitigate these issues, the authors construct a domain-specific knowledge graph and integrate it with the GraphRAG framework to retrieve query-relevant subgraphs as relational evidence. A graph neural network is employed to capture multi-hop dependencies among symptoms, diseases, and control measures. This graph-derived evidence is then incorporated into the input of a ChatGLM backbone model, which is efficiently fine-tuned using LoRA, to guide the generation of domain-consistent recommendations. This work represents the first application of graph-augmented reasoning in tobacco plant protection, significantly improving accuracy and reliability on complex multi-hop and comparative reasoning tasks while effectively suppressing hallucinations and enhancing the professionalism of diagnostic and treatment suggestions.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Reasoning under Uncertainty: Graphical Models
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

This paper proposes a graph-augmented reasoning framework for tobacco pest and disease management that integrates structured domain knowledge into large language models. Building on GraphRAG, we construct a domain-specific knowledge graph and retrieve query-relevant subgraphs to provide relational evidence during answer generation. The framework adopts ChatGLM as the Transformer backbone with LoRA-based parameter-efficient fine-tuning, and employs a graph neural network to learn node representations that capture symptom-disease-treatment dependencies. By explicitly modeling diseases, symptoms, pesticides, and control measures as linked entities, the system supports evidence-aware retrieval beyond surface-level text similarity. Retrieved graph evidence is incorporated into the LLM input to guide generation toward domain-consistent recommendations and to mitigate hallucinated or inappropriate treatments. Experimental results show consistent improvements over text-only baselines, with the largest gains observed on multi-hop and comparative reasoning questions that require chaining multiple relations.

Problem

• tobacco pest and disease management
• large language models
• knowledge graph
• reasoning
• hallucination mitigation

Innovation

• graph-augmented reasoning
• knowledge graph
• large language models
• parameter-efficient fine-tuning
• graph neural network