Poster
in
Workshop: Building Trust in LLMs and LLM Applications: From Guardrails to Explainability to Regulation

ToolScan: A Benchmark For Characterizing Errors In Tool-Use LLMs

Shirley Kokane ⋅ Ming Zhu ⋅ Tulika Awalgaonkar ⋅ Jianguo Zhang ⋅ Akshara Prabhakar ⋅ Thai Hoang ⋅ Zuxin Liu ⋅ Rithesh Ramapura Narasimha Murthy ⋅ Liangwei Yang ⋅ Weiran Yao ⋅ Juntao Tan ⋅ Zhiwei Liu ⋅ Shelby Heinecke ⋅ Huan Wang ⋅ Juan Carlos Niebles ⋅ Caiming Xiong ⋅ Silvio Savarese

Project Page [ OpenReview]

Abstract

Evaluating Large Language Models (LLMs) is one of the most critical aspects of building a performant compound AI system. Since the output from LLMs propagate to downstream steps, identifying LLM errors is crucial to system performance. A common task for LLMs in AI systems is tool use. While there are several benchmark environments for evaluating LLMs on this task, they typically only give a success rate without any explanation of the failure cases. To solve this problem, we introduce ToolScan, a new benchmark to identify error patterns in LLM output on tool-use tasks. Our benchmark data set comprises of queries from diverse environments that can be used to test for the presence of seven newly characterized error patterns. Using ToolScan, we show that even the most prominent LLMs exhibit these error patterns in their outputs. Researchers can use these insights from ToolScan to guide their error mitigation strategies. We open-source our evaluation framework at https://anonymous.4open.science/r/ToolScan-1474 .

Chat is not available.