OpaqueToolsBench: Learning Nuances of Tool Behavior Through Interaction
This addresses the challenge of improving LLM agent performance in real-world tasks with opaque tools, offering a more efficient solution for tool documentation.
The paper tackles the problem of LLM agents struggling with opaque tools that lack clear documentation by introducing OpaqueToolsBench, a benchmark with underspecified tools, and finds that existing documentation methods are expensive and unreliable. They propose ToolObserver, a framework that iteratively refines tool documentation using execution feedback, which outperforms baselines and reduces token usage by 3.5-7.5x in test-time exploration.
Tool-calling is essential for Large Language Model (LLM) agents to complete real-world tasks. While most existing benchmarks assume simple, perfectly documented tools, real-world tools (e.g., general "search" APIs) are often opaque, lacking clear best practices or failure modes. Can LLM agents improve their performance in environments with opaque tools by interacting and subsequently improving documentation? To study this, we create OpaqueToolsBench, a benchmark consisting of three distinct task-oriented environments: general function calling, interactive chess playing, and long-trajectory agentic search. Each environment provides underspecified tools that models must learn to use effectively to complete the task. Results on OpaqueToolsBench suggest existing methods for automatically documenting tools are expensive and unreliable when tools are opaque. To address this, we propose a simple framework, ToolObserver, that iteratively refines tool documentation by observing execution feedback from tool-calling trajectories. Our approach outperforms existing methods on OpaqueToolsBench across datasets, even in relatively hard settings. Furthermore, for test-time tool exploration settings, our method is also efficient, consuming 3.5-7.5x fewer total tokens than the best baseline.