PAC-MCTS: Bias-Aware Pruning for Robust LLM-Guided Search and Planning

TL;DR

PAC-MCTS introduces a bias-aware pruning method for search and planning with large language models, providing formal safety guarantees and improving robustness and efficiency in complex environments.

Contribution

It formulates node expansion as a biased Best-Arm Identification problem, deriving sample complexity bounds and proposing a dynamic, bias-aware pruning framework for LLM-guided search.

Findings

PAC-MCTS reduces API evaluations by up to 78%.

It achieves over 3x higher sample efficiency under strict compute budgets.

Experiments validate robustness improvements with increasing bias.

Abstract

As search depth increases in autonomous reasoning and embodied planning, candidate action spaces expand exponentially, often exhausting computational budgets. While heuristic pruning is a critical countermeasure, existing approaches lack formal safety guarantees when guided by surrogate evaluators such as Large Language Models (LLMs), which exhibit systematic biases. We formulate node expansion as a localized Best-Arm Identification (BAI) problem under bounded bias $L$ and derive a sample complexity upper bound of $\mathcal{O}((\Delta-4L)^{-2})$, identifying $\Delta > 4L$ as the regime where safe elimination is feasible. We further establish an information-theoretic lower bound of $\Omega((\Delta-2L)^{-2})$ that characterizes the structural limits of biased exploration. Motivated by these results, we propose PAC-MCTS, a bias-aware pruning framework that dynamically adapts confidence bounds during search. Experiments on Blocksworld and ALFWorld demonstrate that PAC-MCTS consistently improves robustness and search efficiency over strong pruning baselines, achieving up to 78\% fewer API evaluations and over 3$\times$ higher sample efficiency under strict compute budgets. Ablation studies further validate the predicted degradation behavior as evaluator bias increases.