Reproduce “Supervised Learning Objectives for High-Dimensional Treatment Effects” 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
Nikita Dhawan, Arnav Paruthi, Andrew Kim, Lovedeep Gondara, Jekaterina Novikova, Chris J. Maddison
The authors introduce causal risk minimization (CRM) to estimate average outcomes under treatments, including unseen text. The method learns a treatment-to-APO mapping from weighted outcomes and balances confounder moments selected by outcome structure. On a semi-synthetic Amazon Reviews task, SW-CRM with K=2 reaches relative MAE 0.220 and APO correlation 0.874; this depends on causal assumptions and model-class generalization.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.