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

Researchers have developed ToolTree, a new Monte Carlo tree search-based planning system for LLM agents that improves tool selection and usage through dual-feedback evaluation and bidirectional pruning. The system achieves approximately 10% performance gains over existing methods while maintaining high efficiency across multiple benchmarks.

• →ToolTree introduces Monte Carlo tree search to LLM agent tool planning, moving beyond greedy reactive strategies.
• →The system uses dual-stage LLM evaluation and bidirectional pruning to optimize tool usage trajectories.
• →Testing across 4 benchmarks shows consistent 10% performance improvement over state-of-the-art methods.
• →The approach addresses inter-tool dependencies and enables more informed decision-making in complex multi-step tasks.
• →ToolTree maintains high efficiency while delivering superior performance in both open-set and closed-set planning scenarios.