Suppressing Domain-Specific Hallucination in Construction LLMs: A Knowledge Graph Foundation for GraphRAG and QLoRA on River and Sediment Control Technical Standards

TL;DR

This study compares three methods for answering technical questions from Japan's River and Sediment Control standards using open-source LLMs, finding domain-specific fine-tuning outperforms graph augmentation in accuracy and speed.

Contribution

It demonstrates that domain-specific fine-tuning with QLoRA significantly improves LLM performance over graph-based retrieval methods in technical standard question answering.

Findings

8B QLoRA fine-tuned model outperforms larger baseline and GraphRAG in accuracy.

Fine-tuned models run three times faster than baseline.

Knowledge graph augmentation offers moderate gains but is less effective than fine-tuning.

Abstract

This paper addresses the challenge of answering technical questions derived from Japan's River and Sediment Control Technical Standards -- a multi-volume regulatory document covering survey, planning, design, and maintenance of river levees, dams, and sabo structures -- using open-source large language models running entirely on local hardware. We implement and evaluate three complementary approaches: Case A (plain 20B LLM baseline), Case B (8B LLM with QLoRA domain fine-tuning on 715 graph-derived QA pairs), and Case C (20B LLM augmented with a Neo4j knowledge graph via GraphRAG). All three cases use the Swallow series of Japanese-adapted LLMs and are evaluated on a 100-question benchmark spanning 8 technical categories, judged automatically by an independent LLM (Qwen2.5-14B, score 0--3). The key finding is a performance inversion: the 8B QLoRA fine-tuned model (Case B) achieves a judge average of 2.92/3 -- surpassing both the 20B plain baseline (Case A: 2.29/3, $+$0.63) and the 20B GraphRAG approach (Case C: 2.62/3, $+$0.30) -- while running at 3$\times$ faster latency (14.2s vs. 42.2s for Case A). GraphRAG provides moderate gains ($+$0.33 over baseline) but is outperformed by domain-specific fine-tuning in both quality and efficiency. We document the full engineering pipeline, including knowledge graph construction (200 nodes, 268 relations), QLoRA training data generation from Neo4j relations, training on a single GPU (16 GB VRAM) using unsloth, GGUF Q4_K_M quantisation and Ollama deployment, and the graph retrieval and re-ranking design. High-level engineering lessons are distilled in the main body; implementation pitfalls and toolchain details are documented in Supplementary Materials.