Reproduce “Bandits with Preference Feedback: A Stackelberg Game Perspective” 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

Barna Pásztor, Parnian Kassraie, Andreas Krause

MAXMINLCB selects pairs of candidates from noisy preference feedback, where an evaluator identifies the better option. It models utility differences with a kernel and chooses a leader anticipating a follower’s challenger, using confidence bounds to balance selection and informative comparisons. Across eight benchmark settings, it has lowest regret in four; its guarantee assumes bounded RKHS and a restricted candidate set.

Summary by alphaXiv.

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