Reproduce “Kernel Stein Tests for Multiple Model Comparison” 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
Jen Ning Lim, Makoto Yamada, Bernhard Schölkopf, Wittawat Jitkrittum
The authors extend relative-fit testing to lists of models, selecting the candidate with the smallest estimated discrepancy as a temporary reference and testing which others are worse. One method accounts for reference selection; another splits data and adjusts for multiple tests. In face-generation trials, generated Model 1 was often rejected, while close real mixtures were rarely rejected; these rates do not prove which model is best.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.