Reproduce “Exploration via Exploitation: The Blessing of Reward Diversity in Personalized Federated RL” 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
Jiin Woo, Gauri Joshi, Yuejie Chi
PF-EFVI lets agents in a shared, unknown environment learn separate reward-specific policies using pooled transition data, without explicit exploration bonuses. It relies on diverse rewards to provide collective state-action coverage, required for its guarantee. In synthetic experiments, coverage was absent at M=1 and 4 but positive from M=16, where regret was comparable to PF-UCBVI; this evidence is task-specific.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.