Reproduce “Mitigating Model Drift in Developing Economies Using Synthetic Data and Outliers” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ilyas Varshavskiy, Bonu Boboeva, Shuhrat Khalilbekov, Azizjon Azimi, Sergey Shulgin, Akhlitdin Nizamitdinov, Haitz Sáez de Ocáriz Borde
The paper evaluates whether synthetic tail examples can improve default prediction after distribution shifts. Its pipeline adds Gaussian outliers beyond 3 standard deviations to synthetic tabular data, while models using real/synthetic mixtures are assessed before and after shocks with stability scores. On Azerbaijan data, TabPFN’s Stabilization Uplift was 0.9981 with 5% outliers versus 0.9827 without outliers, but effects varied by dataset, model, and share.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.