ToolScan: A Benchmark for Characterizing Errors in Tool-Use LLMs
This addresses the need for better error analysis in LLM evaluation for researchers and developers building AI systems, though it is incremental as it builds on existing benchmarks by adding error characterization.
The paper tackles the problem of evaluating Large Language Models (LLMs) in tool-use tasks by introducing TOOLSCAN, a benchmark that identifies and characterizes seven error patterns in LLM outputs, showing that even prominent LLMs exhibit these errors.
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.