ToolTree: Efficient LLM Agent Tool Planning via Dual-Feedback Monte Carlo Tree Search and Bidirectional Pruning

TL;DR

ToolTree introduces a Monte Carlo tree search-inspired framework with dual-feedback and bidirectional pruning for efficient, foresightful tool planning in LLM agents, significantly outperforming existing methods across multiple benchmarks.

Contribution

It proposes a novel planning paradigm for LLM agents that incorporates dual-feedback evaluation and bidirectional pruning, enabling more informed and efficient tool sequence planning.

Findings

Achieves around 10% performance improvement over state-of-the-art methods.

Demonstrates effectiveness on 4 diverse benchmarks.

Maintains high efficiency while improving planning accuracy.

Abstract

Large Language Model (LLM) agents are increasingly applied to complex, multi-step tasks that require interaction with diverse external tools across various domains. However, current LLM agent tool planning methods typically rely on greedy, reactive tool selection strategies that lack foresight and fail to account for inter-tool dependencies. In this paper, we present ToolTree, a novel Monte Carlo tree search-inspired planning paradigm for tool planning. ToolTree explores possible tool usage trajectories using a dual-stage LLM evaluation and bidirectional pruning mechanism that enables the agent to make informed, adaptive decisions over extended tool-use sequences while pruning less promising branches before and after the tool execution. Empirical evaluations across both open-set and closed-set tool planning tasks on 4 benchmarks demonstrate that ToolTree consistently improves performance while keeping the highest efficiency, achieving an average gain of around 10\% compared to the state-of-the-art planning paradigm.