When2Call: When (not) to Call Tools

2025/04/26 by Hayley Ross, Ross, Hayley, Ameya Sunil Mahabaleshwarkar +3 · 11 citations
Computer Science · Medicine · #Artificial Intelligence in Healthcare and Education #Computation and Language (cs.CL) #FOS: Computer and information sciences #Natural Language Processing Techniques #Topic Modeling

paper · pdf · doi:10.48550/arxiv.2504.18851

openalex publication_date 2025/04/26 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Leveraging external tools is a key feature for modern Language Models (LMs) to expand their capabilities and integrate them into existing systems. However, existing benchmarks primarily focus on the accuracy of tool calling -- whether the correct tool is called with the correct parameters -- and less on evaluating when LMs should (not) call tools. We develop a new benchmark, When2Call, which evaluates tool-calling decision-making: when to generate a tool call, when to ask follow-up questions and when to admit the question can't be answered with the tools provided. We find that state-of-the-art tool-calling LMs show significant room for improvement on When2Call, indicating the importance of this benchmark. We also develop a training set for When2Call and leverage the multiple-choice nature of the benchmark to develop a preference optimization training regime, which shows considerably more improvement than traditional fine-tuning. We release the benchmark and training data as well as evaluation scripts at https://github.com/NVIDIA/When2Call.

Cited by

• From Proof to Program: Characterizing Tool-Induced Reasoning Hallucinations in Large Language Models
• ToolMind Technical Report: A Large-Scale, Reasoning-Enhanced Tool-Use Dataset
• Structured Uncertainty guided Clarification for LLM Agents
• SCRIBE: Structured Chain Reasoning for Interactive Behaviour Explanations using Tool Calling
• When Top-K Misses the Decision: Tool-Call Drift in Multi-Teacher On-Policy Distillation
• Temporal Blindness in Multi-Turn LLM Agents: Misaligned Tool Use vs. Human Time Perception
• Foundational Automatic Evaluators: Scaling Multi-Task Generative Evaluator Training for Reasoning-Centric Domains
• RAGCap-Bench: Benchmarking Capabilities of LLMs in Agentic Retrieval Augmented Generation Systems
• Therefore I am. I Think
• COMPASS: Benchmarking Constrained Optimization in LLM Agents
• NVIDIA Nemotron Nano 2: An Accurate and Efficient Hybrid Mamba-Transformer Reasoning Model

Related

• Achieving Tool Calling Functionality in LLMs Using Only Prompt Engineering Without Fine-Tuning
• Learning to Ask: When LLM Agents Meet Unclear Instruction
• SMARTCAL: An Approach to Self-Aware Tool-Use Evaluation and Calibration
• CRITICTOOL: Evaluating Self-Critique Capabilities of Large Language Models in Tool-Calling Error Scenarios
• ToolTalk: Evaluating Tool-Usage in a Conversational Setting