Reproduce “Contextual Multi-armed Bandit Algorithm for Semiparametric Reward Model” 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
Gi-Soo Kim, Myunghee Cho Paik
The authors propose a contextual bandit for article selection with feedback only for the chosen article and a shared baseline that may vary across visits. The method centers the chosen article’s context against a probability-weighted average of available contexts and uses Thompson sampling. In Yahoo! replay, it estimates 4.4% more clicks than linear Thompson sampling; this is not a live result.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.