Salvato in:
Dettagli Bibliografici
Autori principali: Hamad, Hassan, Xu, Yingru, Zhao, Liang, Yan, Wenbo, Gyanchandani, Narendra
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2510.17052
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912659576717312
author Hamad, Hassan
Xu, Yingru
Zhao, Liang
Yan, Wenbo
Gyanchandani, Narendra
author_facet Hamad, Hassan
Xu, Yingru
Zhao, Liang
Yan, Wenbo
Gyanchandani, Narendra
contents Tool-augmented large language models (LLMs) are increasingly employed in real-world applications, but tool usage errors still hinder their reliability. We introduce ToolCritic, a diagnostic framework that evaluates and improves LLM behavior in multi-turn, tool-augmented dialogues. ToolCritic detects eight distinct error types specific to tool-calling (e.g., premature invocation, argument misalignment, and misinterpretation of tool outputs) and provides targeted feedback to the main LLM. The main LLM, assumed to have strong reasoning, task understanding and orchestration capabilities, then revises its response based on ToolCritic's feedback. We systematically define these error categories and construct a synthetic dataset to train ToolCritic. Experimental results on the Schema-Guided Dialogue (SGD) dataset demonstrate that ToolCritic improves tool-calling accuracy by up to 13% over baselines, including zero-shot prompting and self-correction techniques. This represents a promising step toward more robust LLM integration with external tools in real-world dialogue applications.
format Preprint
id arxiv_https___arxiv_org_abs_2510_17052
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ToolCritic: Detecting and Correcting Tool-Use Errors in Dialogue Systems
Hamad, Hassan
Xu, Yingru
Zhao, Liang
Yan, Wenbo
Gyanchandani, Narendra
Artificial Intelligence
Tool-augmented large language models (LLMs) are increasingly employed in real-world applications, but tool usage errors still hinder their reliability. We introduce ToolCritic, a diagnostic framework that evaluates and improves LLM behavior in multi-turn, tool-augmented dialogues. ToolCritic detects eight distinct error types specific to tool-calling (e.g., premature invocation, argument misalignment, and misinterpretation of tool outputs) and provides targeted feedback to the main LLM. The main LLM, assumed to have strong reasoning, task understanding and orchestration capabilities, then revises its response based on ToolCritic's feedback. We systematically define these error categories and construct a synthetic dataset to train ToolCritic. Experimental results on the Schema-Guided Dialogue (SGD) dataset demonstrate that ToolCritic improves tool-calling accuracy by up to 13% over baselines, including zero-shot prompting and self-correction techniques. This represents a promising step toward more robust LLM integration with external tools in real-world dialogue applications.
title ToolCritic: Detecting and Correcting Tool-Use Errors in Dialogue Systems
topic Artificial Intelligence
url https://arxiv.org/abs/2510.17052