cs.AIMar 3, 2026

GSI Agent: Domain Knowledge Enhancement for Large Language Models in Green Stormwater Infrastructure

Authors: Shaohuang Wang

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Green Stormwater Infrastructure (GSI) systems, such as permeable pavement, rain gardens, and bioretention facilities, require continuous inspection and maintenance to ensure long-term performance. However, domain knowledge about GSI is often scattered across municipal manuals, regulatory documents, and inspection forms. As a result, non-expert users and maintenance staff may struggle to obtain reliable and actionable guidance from field observations. Although Large Language Models (LLMs) have demonstrated strong general reasoning and language generation capabilities, they often lack domain-specific knowledge and may produce inaccurate or hallucinated answers in engineering scenarios. This limitation restricts their direct application to professional infrastructure tasks. In this paper, we propose GSI Agent, a domain-enhanced LLM framework designed to improve performance in GSI-related tasks. Our approach integrates three complementary strategies: (1) supervised fine-tuning (SFT) on a curated GSI instruction dataset, (2) retrieval-augmented generation (RAG) over an internal GSI knowledge base constructed from municipal documents, and (3) an agent-based reasoning pipeline that coordinates retrieval, context integration, and structured response generation. We also construct a new GSI Dataset aligned with real-world GSI inspection and maintenance scenarios. Experimental results show that our framework significantly improves domain-specific performance while maintaining general knowledge capability. On the GSI dataset, BLEU-4 improves from 0.090 to 0.307, while performance on the common knowledge dataset remains stable (0.304 vs. 0.305). These results demonstrate that systematic domain knowledge enhancement can effectively adapt general-purpose LLMs to professional infrastructure applications.

Figures & tables

CategoryTypical methodKey idea and trade-off
Dynamic injectionRAGRetrieve external documents at inference time; easy to update, but depends on retrieval quality.
Static injectionSFTPut domain patterns into model parameters; strong for style/tasks, but harder to update.
Model adapterLoRATrain small rank adapters; efficient and low-cost, but capacity is limited.
Prompt optimizationPromptControl behavior without changing weights; fast but may be brittle.
Table 1: Knowledge injection taxonomy used in this paper.
Figure 1: Architecture of the proposed GSI knowledge-enhanced LLM system integrating domain-specific fine-tuning, retrieval-augmented generation, and agent-based reasoning.
FieldDescription
_idUnique identifier (UUIDv4).
_sourceSource document (PDF) providing reference information.
_source_locationGeographical tag (e.g., “Philadelphia, PA”) or empty if not location-specific.
_task_typeOne of nine predefined task families.
_deployment_typeIntended usage: fine-tuning or rag .
_created_atTimestamp of record creation (UTC, RFC3339).
Table 2: Schema of GSI Dataset for instruction-based fine-tuning.
LocationCountPercentage
Philadelphia, PA521954.2%
None479145.8%
Table 3: Source location distribution in GSI Dataset.
TypeCountPercentage
Fine-tuning746073.3%
RAG349526.7%
Table 4: Deployment type distribution in GSI Dataset.
Task TypeCountPercentage
Question Answering530031.2%
Verification / Judgment524130.9%
Generation / Composition257315.2%
Information Extraction172410.1%
Classification12257.2%
Reasoning / Math / Logic7584.4%
Table 5: Task type distribution in GSI Dataset.
Figure 2: Task type distribution in GSI Dataset.
Task TypeCountPercentage
Question Answering40040%
Classification30030%
Reasoning / Logic30030%
Table 6: Summary of Common Knowledge Dataset used to evaluate general knowledge retention.
MetricLevelWhat it measures
BLEU-4LexicalN-gram precision with brevity penalty; good for short factual text.
ROUGE-1/2/LLexicalRecall-style overlap; captures coverage for summary-like answers.
Micro-F1LabelAggregated F1 for classification-style tasks (e.g., issue type).
Sentence-BERTSemanticEmbedding cosine similarity; complements lexical overlap.
G-Eval (LLM Judge)Semantic/LogicLLM-based scoring for correctness and coherence.
Human ExpertRealExpert rating on usefulness and correctness (small sample).
Table 7: Evaluation metrics and their roles.
BaselineRAGSFTAgentNotes
Base LLM×\times×\times×\timesDirect prompting on base model.
Base LLM + RAG✓×\times×\timesRetrieval improves factuality; no parameter updates.
Fine-tuned LLM + RAG✓✓✓Full system with LoRA-SFT, RAG, and agent reasoning.
Table 8: Baselines used in evaluation.
SettingValue
finetuning_typeLoRA
bf16true
templateqwen3_vl
lora_alpha16
lora_dropout0
lora_rank8
Table 9: LoRA fine-tuning configuration.
MetricCommon Knowledge DatasetGSI Dataset
Base LLMGSI LLMBase LLMGSI LLM
BLEU-40.3040.3050.0900.307
ROUGE-10.3520.3510.1570.204
ROUGE-20.1460.1460.0320.111
ROUGE-L0.2230.2230.0710.153
Sentence-BERT0.8610.8690.5440.742
Table 10: Main results on general and domain datasets.
MethodG-Eval Score
LLM + RAG0.51
LLM + Fine-tuning0.63
LLM + RAG + Fine-tuning0.72
Table 11: Ablation study of knowledge-injection strategies.
Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

#TitleYearLink / Notes
1Stormwater Management Guidance Manual2023https://water.phila.gov/wp-content/uploads/files/stormwater-management-guidance-manual.pdf
2Pennsylvania Stormwater BMPs Manual–https://greenport.pa.gov/elibrary/GetFolder?FolderID=1368916
3City of Philadelphia Green Streets Design Manual2014https://www.phila.gov/media/20160504172218/Green-Streets-Design-Manual-2014.pdf
4Green Stormwater Infrastructure Maintenance Manual2016https://water.phila.gov/wp-content/uploads/GSI-Maintenance-Manual_v2_2016.pdf
5Green City, Clean Waters Plan2009https://www.phila.gov/media/20160421133948/green-city-clean-waters.pdf
6Green City, Clean Waters Partnership Agreement2012https://water.phila.gov/wp-content/uploads/files/EPA_Partnership_Agreement.pdf
Appendix
Table 14
More in this topic

Knowledge Augmentation for Language Models

Knowledge augmentation for language models adds external or newly learned information to improve factuality, reasoning, and domain performance.

See all 178 papers

Explore similar work

CardsList
Mar 3, 2026

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

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.

Takato Yasuno

Question AnsweringGraph Retrieval-Augmented GenerationParameter-Efficient Fine-Tuning
AskPDF
Apr 9, 2026

Vision-Language Foundation Models for Comprehensive Automated Pavement Condition Assessment

General-purpose vision-language models demonstrate strong performance in everyday domains but struggle with specialized technical fields requiring precise terminology, structured reasoning, and adherence to engineering standards. This work addresses whether domain-specific instruction tuning can enable comprehensive pavement condition assessment through vision-language models. PaveInstruct, a dataset containing 278,889 image-instruction-response pairs spanning 32 task types, was created by unifying annotations from nine heterogeneous pavement datasets. PaveGPT, a pavement foundation model trained on this dataset, was evaluated against state-of-the-art vision-language models across perception, understanding, and reasoning tasks. Instruction tuning transformed model capabilities, achieving improvements exceeding 20% in spatial grounding, reasoning, and generation tasks while producing ASTM D6433-compliant outputs. These results enable transportation agencies to deploy unified conversational assessment tools that replace multiple specialized systems, simplifying workflows and reducing technical expertise requirements. The approach establishes a pathway for developing instruction-driven AI systems across infrastructure domains including bridge inspection, railway maintenance, and building condition assessment.

Blessing Agyei Kyem, Joshua Kofi Asamoah, Anthony Dontoh +1

Department of Civil, Construction and Environmental Engineering, North Dakota State University, Fargo, 58102, ND, USA · Department of Civil, Construction and Environmental Engineering, University of Memphis, Memphis, 38152, TN, USA

Multimodal Instruction TuningVLM Evaluation
AskPDF
Jun 27, 2026

Customized Generative AI Agent for Transportation Engineering Practice: A Development and Continued Pre-training Guideline

Recent advancements in generative artificial intelligence (AI) and large language models (LLMs) have shown significant promise in automating complex reasoning, summarization, and question-answering tasks. However, the effectiveness of general-purpose LLMs in specialized engineering domains remains limited due to insufficient exposure to technical standards, engineering terminology, and domain-specific semantics. This study proposes a systematic approach to developing a customized generative AI agent for transportation engineering applications. A curated corpus of U.S. transportation manuals, design guidelines, and regulatory documents is used to conduct continued pretraining of six state-of-the-art LLMs through a unified low-rank adaptation (LoRA) framework. The training process is monitored to ensure convergence and model stability. Performance is evaluated using standard natural language processing metrics, including BLEU-4 and ROUGE, with Qwen2.5-7B and LLaMA-3.1-8B demonstrating the highest domain alignment and response quality. Results validate the effectiveness of LoRA-based adaptation in improving LLM performance on technical content interpretation and context-specific reasoning. This work contributes a reproducible development framework for constructing domain-specialized generative AI agents, supporting broader deployment in transportation research, design, planning, and policy analysis.

Dianwei Chen, Yuan-Zheng Lei, Zifan Zhang +2

Domain-Adaptive PretrainingLLM Agent Training
AskPDF