Reproduce “Simultaneous AlphaZero: Extending Tree Search to Markov Games” 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
Tyler Becker, Zachary Sunberg
Simultaneous AlphaZero (SM-AZ) enables bounded search in deterministic, two-player, zero-sum Markov games where players choose actions simultaneously. Learned value estimates and regret and strategy priors guide the search. In satellite custody cross-play, at 50 tree queries and transfer weight 0.025, it has a seat-balanced utility advantage of 0.73 ± 0.30 over its value-only variant; this single-model result is not a Nash gap.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.