Large Language Models and Geoscience Transformers for Predictive Mapping of Canadian Critical Minerals
Bibliographic record
Abstract
Abstract Data-driven MPM (mineral prospectivity mapping) of critical minerals, which are elements or minerals with strategic importance and high supply chain risk, is vital for national land use planning. Recently, MPM has been practiced using supervised machine learning classification algorithms. However, applying such algorithms to Canadian critical minerals presents two major challenges. The first stems from the nature of geological knowledge, which is primarily stored in unstructured text. However, most supervised machine learning algorithms struggle to directly incorporate this textual information into predictive models. The second challenge arises from the limited number of known mineral deposits associated with many critical minerals in Canada, resulting in insufficient training labeled data for supervised classification tasks. To address the first challenge, this study employed natural language processing (NLP) techniques and large language models (LLMs) to extract and transform geoscientific knowledge embedded in geoscience text corpora into predictive features for MPM. LLMs operate based on transformer deep learning architectures that use self-attention mechanisms to capture contextual relationships within natural language. A domain-specific LLM, which was fine-tuned in this study and evaluated using geology-related inquiries, was employed for MPM. To address the second challenge, a separate transformer model was developed using a self-supervised learning approach that integrates diverse geophysical, geochronological, and textual data, eliminating the dependency on a substantial number of labeled training samples. The prospectivity model generated using the proposed transformer model significantly reduced the search space—by an average of 87%—for the targeted type of mineral deposits. The findings of this study demonstrate the effectiveness of transformer-based architectures and LLMs in overcoming key limitations of modern MPM approaches for critical mineral exploration.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 teacher head, 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".