BONSAI

Software

Selected software

1. MLXP

   Launch experiment sweeps, record configurations, code versions and artifacts, and query results in one Python workflow. MLXP supports cluster job submission and makes experiments easier to reproduce. ACM REP 2024.

   Code Documentation ArXiv
2. EquiTabPFN

   A tabular foundation model whose predictions respect class-label permutations, avoiding costly permutation ensembling and accommodating more classes than seen during pretraining. Includes training and benchmark code. NeurIPS 2025.

   Code ArXiv
3. funcBO

   Functional bilevel optimization with neural-network inner models, without requiring strong convexity in their parameters. Includes instrumental-regression experiments and baselines to reproduce and extend the method. NeurIPS 2024 · Spotlight.

   Code ArXiv