Plus: we propose a planning-augmented sampling framework (plus) that integrates Monte Carlo Tree Search (MCTS) with Polynomial Upper Confidence Trees (PUCT) and a tunable alpha-greedy sampling mechanism