Reproduce “Data-Efficient Scaling-Law Discovery for Recommendation Systems via Coupled Hyperparameter Scaling” 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

John Martabano Landy, Louis C. Tiao, Yao Xuan, Luc Kien Hang, Rupinder P. Khandpur, Arvind M. Vepa, Srinidhi Raghavendran, Anirudh Agrawal and 5 more

The authors introduce Axia, which helps allocate serving compute across ranking and recommendation models by searching compact architecture controls. It profiles inference costs, then uses uncertainty-aware estimates to select full-training trials and build a quality–cost frontier. Across 34 model iterations, mean offline normalized-entropy improvement was 0.66%, and cumulative online lift summed across disjoint model sets was 0.36%; deployment comparisons were observational.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.