Reproduce “Monte Carlo Tree Search with Boltzmann Exploration” 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
Michael Painter, Mohamed Baioumy, Nick Hawes, Bruno Lacerda
Boltzmann Tree Search (BTS) separates exploration from recommendation in Monte Carlo tree search. It samples actions stochastically using reward estimates but bases recommendations on reward-only Bellman values; Decaying Entropy Tree Search (DENTS) adds an entropy bonus to guide search while keeping that recommendation. Both have consistency guarantees; in Go, DENTS beat PUCT 58–42, but not at equal trial counts.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.