Reproduce “What preferences can - and cannot - predict in multi-agent online learning” 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
Omar Abbadi, Rida Laraki, Panayotis Mertikopoulos
The paper examines how preference graphs constrain long-run behavior under follow-the-regularized-leader (FTRL) learning. Players use accumulated payoff scores and a regularizer to update their randomized strategies. A three-player, two-action counterexample shows that a preference-closed set can be unstable, while closed subgames are asymptotically stable under FTRL. Payoff-based resilience conditions provide sufficient, but not necessary, stability guarantees.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.