GATE: Graph-based Adaptive Tool Evolution Across Diverse Tasks

TL;DR

GATE introduces a graph-based, adaptive framework for evolving a reusable tool library across diverse tasks in LLM-enabled environments. By coordinating two agents—the Task Solver and Tool Manager—and employing GraphRank retrieval on a hierarchical Tool Graph, GATE dynamically creates, merges, and prunes tools to balance quantity, complexity, and utility. The approach delivers substantial gains in open-ended environments (e.g., Minecraft) and improves performance across code generation and agent tasks, while demonstrating strong zero-shot generalization to unseen tasks. The findings highlight the value of tool invocation networks and adaptive evolution in scalable, multi-task tool-making for autonomous agents.

Abstract

Large Language Models (LLMs) have shown great promise in tool-making, yet existing frameworks often struggle to efficiently construct reliable toolsets and are limited to single-task settings. To address these challenges, we propose GATE (Graph-based Adaptive Tool Evolution), an adaptive framework that dynamically constructs and evolves a hierarchical graph of reusable tools across multiple scenarios. We evaluate GATE on open-ended tasks (Minecraft), agent-based tasks (TextCraft, DABench), and code generation tasks (MATH, Date, TabMWP). Our results show that GATE achieves up to 4.3x faster milestone completion in Minecraft compared to the previous SOTA, and provides an average improvement of 9.23% over existing tool-making methods in code generation tasks and 10.03% in agent tasks. GATE demonstrates the power of adaptive evolution, balancing tool quantity, complexity, and functionality while maintaining high efficiency. Code and data are available at \url{https://github.com/ayanami2003/GATE}.

Figures (17)

• Figure 1: Performance of GATE in Minecraft. GATE continually discovers new Minecraft items and skills during exploration, significantly outperforming other methods.
• Figure 2: GATE consists of two agents: the Task Solver and the Tool Manager, which interact with an Adaptive Tool Graph. Key processes include Tool Requirement, Generation, Pruning, Merging, and Tool Graph Updates.
• Figure 3: Map coverage: bird’s eye views of Minecraft maps.
• Figure 4: Zero-shot Generalization on Unseen Tasks. The figure visualizes the intermediate progress of each method on two tasks. See Figure \ref{['fig:diamon and compass']} for the other two tasks. ReAct and Reflexion are excluded from the plot due to their lack of meaningful progress.
• Figure 5: Evolution of the tool graph. We visualize the progression of the tool graph in the Minecraft task, capturing snapshots every 40 steps. The complete evolution for other tasks is provided in the Appendix \ref{['subsec:tool-graph']}. For clarity, basic tools are excluded from the visualization, as they are generally connected to tools at every level.