Code to replicate the generation and LLM-as-a-judge evaluation of the FactEHR dataset.
- I. Installation
- II. Data Dependencies
- III. Experiment Pipeline [WIP]
- IV. Experiment Runtimes & Costs
Use editable mode during development.
python -m pip install -e .
To run all unit tests in tests/
run this from the project root
pytest
See detailed overview of the FactEHR dataset.
See here for summary of release files.
See experiments for an overview of the experiment pipeline.