SlopeBench

SlopeBench is an open-source Python package for reproducible slope-stability surrogate benchmarking and false-safe diagnostics.

What the software does

SlopeBench validates sample solver-label data, tracks solver-label provenance, summarizes public benchmark recovery records, evaluates feature-complete transfer cases, runs physical-consistency checks, computes conformal false-safe diagnostics, records negative-transfer warnings, and generates reproducible reports.

What the software does not do

• SlopeBench is not a final design tool.
• It does not replace project-specific LEM/FEM analysis.
• It does not provide field validation.
• It does not claim proprietary commercial software validation unless such runs are explicitly supplied by the user.
• It does not certify slope safety.
• It supports reproducible research, audit, teaching, and preliminary screening diagnostics.

Installation

python -m pip install -e .

Quickstart

slopebench reproduce-sample --out outputs/sample_run

Command-line usage

slopebench --help
slopebench validate-data
slopebench summarize
slopebench benchmark-recovery
slopebench transfer-check
slopebench physical-check
slopebench conformal
slopebench negative-transfer
slopebench make-report
slopebench reproduce-sample

Python API example

from slopebench.data import load_sample_bundle
from slopebench.conformal import conformal_summary

bundle = load_sample_bundle()
metrics = conformal_summary(bundle["conformal"])
print(metrics)

Data schema

Sample data are stored in data/sample/. Metadata and schema files are stored in data/metadata/.

Reproducing the SoftwareX examples

Run python examples/reproduce_sample_results.py or slopebench reproduce-sample.

Benchmark recovery workflow

slopebench benchmark-recovery summarizes counted, rejected, attempted, and feature-complete public benchmark records.

Feature-complete transfer diagnostics

slopebench transfer-check evaluates only records whose required feature schema is complete.

Physical-consistency checks

slopebench physical-check reports monotonic perturbation violations and preserves the checked direction.

Conformal false-safe diagnostics

slopebench conformal computes residual quantiles, empirical coverage, and false-safe rates near FS thresholds.

Negative-transfer diagnostics

slopebench negative-transfer reports schema and label-provenance warnings for unrelated tabular data.

Output reports

slopebench make-report writes Markdown, HTML, and optional DOCX summaries under the output directory.

Repository structure

The repository contains src/, tests/, examples/, docs/, data/sample/, data/metadata/, manuscript/, figures/, tables/, validation/, and outputs/.

Tests

python -m pytest tests -q

Zenodo archive

Reproducibility archive: https://doi.org/10.5281/zenodo.20516102

Citation

Cite the SoftwareX article, CITATION.cff, and the Zenodo archive.

License

MIT License.

Limitations

SlopeBench is intended for reproducible research, teaching, audit, and preliminary screening diagnostics. It depends on the quality and completeness of supplied solver-label and benchmark records.

Contact

Hongzhi Lu, Universiti Sains Malaysia, Penang, Malaysia. Email: lhz1076277065@gmail.com.