FrontierTrials welcomes focused changes that make manual model evaluation more transparent.
- Minimal import, allocation, integrity, or report bugs.
- Statistical corrections with references and tests.
- Accessibility and offline-packet improvements.
- Better leakage diagnostics that do not rewrite responses.
- Schema proposals grounded in a real evaluation workflow.
Do not submit private prompts, subscription exports, personal data, provider credentials, or copyrighted material you cannot redistribute.
git clone https://github.com/CAOShurong/frontiertrials.git
cd frontiertrials
python -m pip install -e .
python -m unittest discover -s tests -v
python -m pip install ruff build
ruff check src tests
ruff format --check src tests
python -m build- Keep one behavioral purpose per pull request.
- Add tests for every changed outcome.
- Update schemas and methodology docs for durable-format changes.
- Describe any effect on blinding, order balance, privacy, or interpretation.
- Preserve deterministic output and Windows/Linux portability.
By participating, you agree to the Code of Conduct.