RAG pipeline for private well contamination guidance: A comparative study of retrieval and generation strategies
Bibliographic record
Abstract
Access to safe drinking water remains a fundamental public health priority, particularly in rural and semi-urban areas where private wells are a primary source but often lack proper monitoring. This exposes users to microbiological risks such as E.coli and coliform bacteria. Although large language models (LLMs) hold promise in delivering accessible guidance, their performance in specialized low-resource domains remains limited. In this study, we develop a domain-adapted Retrieval-Augmented Generation (RAG) system tailored to support private well owners with contamination concerns. Starting from a naive RAG baseline, we explore key enhancements, including embedding model fine-tuning (BGE-M3) using synthetic QA pairs, query rewriting, and an adaptive reranking technique. Evaluation combines LLM-as-judge metrics via the deepeval framework, statistical significance testing, and expert review of the generated answers. Adaptive reranking with Llama delivered the highest performance (86.34% answer relevancy, 91.6% faithfulness), improved contextual relevancy, and received the highest expert-rated technical accuracy, demonstrating its advantage in factual correctness.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.005 | 0.018 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.006 | 0.004 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".