GeoLLM Qwen3-1.7B v2 — compact knowledge-graph-grounded geotechnical assistant

A Qwen3-1.7B fine-tune that answers undergraduate geotechnical engineering questions from facts supplied to it at runtime, citing the textbook and page each claim came from. Same training recipe as the 4B model, on a smaller base — 1.8 GB, noticeably faster, and correspondingly weaker.

Built for the QAA-funded GeoLLM project (Cardiff, Manchester, Surrey, Glasgow), which is developing a source-grounded alternative to general-purpose chatbots for engineering students.

Which one should you use?

	This model (1.7B)	4B v2
Size	1.8 GB	4.3 GB
Speed	Fast — runs comfortably on modest hardware	Slower
Answer quality	Weaker; shorter, more likely to miss nuance	Recommended for real use

Choose this one when hardware is constrained, or for latency-sensitive demos. Otherwise prefer 4B.

Important: this model expects retrieved facts, not bare questions

It was trained on RAG-shaped examples — the user turn contains the question plus a block of knowledge-graph facts, each ending with its source and page. It learned how to use supplied evidence, not to recall soil mechanics from memory.

Ask it a bare question and it will still answer, but the citation and refusal behaviour it was trained for only appears when you supply facts. The expected prompt shape:

What are the steps of the cone penetrometer liquid limit test?

## Knowledge-graph facts (retrieved subgraph)
- cone penetrometer procedure -HAS_STEP-> Prepare a cup of very moist soil (ProcedureStep) [order=1] (Barnes p.60)
- cone penetrometer procedure -HAS_STEP-> Allow the cone to fall freely into the cup (ProcedureStep) [order=2] (Barnes p.60)
- liquid limit cone method -GOVERNED_BY-> BS 5930:1999 (Standard) (Barnes p.62)

## Sources available
- [Barnes] Graham Barnes — Soil Mechanics: Principles and Practice (2016)

The full retrieval stack that produces this — knowledge graph, hybrid search, grounding checks and a student-facing web UI — is open source and runs with one command:

https://github.com/suhasdevmane/QAA-CEP-GeoLLM

git clone https://github.com/suhasdevmane/QAA-CEP-GeoLLM
cd QAA-CEP-GeoLLM/kg && cp .env.example .env    # set NEO4J_PASSWORD
docker compose up -d                             # GUI at http://localhost:8008

What it was trained to do

Answer only from the supplied facts — and say plainly when they do not cover the question
Cite every claim inline as [Barnes p.60], taken from the fact it used
Challenge false premises — reject physically impossible inputs (a dry mass greater than the wet mass, D60 smaller than D10, saturation above 100%) instead of computing from them
Show working on calculations, and preserve step order in procedures

Training


Base	Qwen/Qwen3-1.7B
Method	QLoRA (4-bit, rank 16), ~22 min on a single 16 GB laptop GPU
Data	456 instruction examples generated from a curated geotechnical knowledge graph (40.6k facts from 8 textbooks, every fact carrying source + page), balanced across five question types, plus synthetic refusal examples
Contamination guard	Every generated question was similarity-checked against the project’s 300-question evaluation benchmark and dropped if close — the benchmark stays a clean held-out test

Evaluation

Scored on a 30-question sample of the project’s EN3300 benchmark (one per lab test × question type), with identical retrieval for every model compared:

Metric	Result
Semantic similarity vs gold answers	0.736 (30-question sample)
Fine-tuning gain over the Qwen3-1.7B base	0.673 → 0.736

The 4B model was additionally evaluated across all 300 questions; this model was not, so treat its figure as indicative rather than directly comparable. Methodology: training/BAKEOFF_RESULTS.md

Limitations

Weaker than the 4B model — that is the trade for size and speed. Prefer 4B unless constrained.
Not a general geotechnical oracle. Without supplied facts it behaves like a small base model.
Scoped to soil-mechanics laboratory testing — the graph is thin on particle-size practice (sieve calibration, sedimentation pretreatment) and this shows in answers on those topics.
Citations should be verified. It is a study aid, not an authority — check the cited page.
Undergraduate teaching context only. Not validated for professional engineering design.

Licence and provenance

Base model Qwen3-1.7B (Apache 2.0). Training data was generated from a knowledge graph distilled from copyrighted geotechnical textbooks; content reaches the model only as re-expressed facts with provenance attached, consistent with the project’s IP position. Released as a research artifact for education and evaluation.

Part of a QAA Collaborative Enhancement Project across Cardiff, Manchester, Surrey and Glasgow. Recommended model: SuhasDevmane55/geollm-qwen3-4b-v2.

Models

Readme