badr 0.1.1

Badr is a framework for training fair machine-learning models that minimize a chosen fairness metric while preserving Pareto-optimal performance across sensitive groups.

BADR - Bilevel Adaptive Rescalarization

Fairness-Informed Pareto Optimization

badr is a Python package that transforms a large range of estimators into fair and Pareto-efficient estimators.

Have a look at badr documentation!

Citations

If you find this repository useful, or you use it in your research, please consider citing the following paper:

@article{tanji2026fairness,
  title   = {Fairness-informed Pareto Optimization: An Efficient Bilevel Framework},
  author  = {Tanji, Sofiane and Vaiter, Samuel and Laguel, Yassine},
  journal = {arXiv preprint arXiv:2601.13448},
  year    = {2026}
}