Reproduce “Beyond Interaction Capacity: Estimator Scaling with Recursive Models for CTR Prediction” 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
Shivang Chopra, Fotis Iliopoulos, Zsolt Kira, Gaurav Menghani
RECAP scales click-through-rate prediction by aggregating related estimators within a compact recursive model, rather than only enlarging one predictor or deploying several full models. It combines signals from independent runs, training checkpoints, and inference routes. On Criteo, RECAP reaches 81.764% test AUC versus 81.771% for a five-model FCN ensemble, with about one fifth its deployed parameters.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.