AsyncTool: Evaluating the Asynchronous Function Calling Capability under Multi-Task Scenarios

2026-05-27

AI Summary

This work addresses the challenge that current large language model agents struggle to effectively manage tool invocation delays in concurrent multi-task settings, leading to inefficient task coordination. To tackle this issue, the authors propose AsyncTool—the first evaluation framework specifically designed for asynchronous tool usage—alongside a novel multi-scenario asynchronous multi-task dataset. The framework introduces efficiency-oriented metrics and a hybrid data evolution strategy, enabling multi-granularity assessment at the step, subtask, and task levels. Experimental results demonstrate a significant performance degradation in existing models under delayed feedback, while models with stronger task coordination capabilities exhibit markedly better robustness. These findings uncover critical failure modes and offer valuable insights for the future design of intelligent agents operating in asynchronous environments.

Abstract

Large language model (LLM)-based agents have shown strong capabilities in using external tools to solve complex tasks. However, existing evaluations often overlook the temporal dimension of tool use, especially the impact of tool response latency, and are usually limited to single-task settings. In real-world applications, multiple tasks often need to be executed concurrently, and overall efficiency depends on whether an agent can use idle time while waiting for tool responses. We refer to this capability as asynchronous tool calling. To evaluate it, we propose AsyncTool, a benchmark for assessing LLM-based agents in interactive multi-task tool-use environments with delayed tool feedback. AsyncTool presents multiple heterogeneous tasks simultaneously and simulates realistic tool response latency during execution. Using a hybrid data evolution strategy, we construct a diverse asynchronous multitasking dataset that covers multiple scenarios and tool-use patterns. We evaluate models at the step, sub-task, and task levels, and introduce efficiency-oriented metrics to measure task coordination and completion efficiency. Extensive experiments show that delayed tool feedback poses substantial challenges to current agents and leads to clear performance degradation. Models that better coordinate task switching, dependency tracking, and state maintenance achieve stronger performance on AsyncTool. Our analysis identifies key failure modes of current tool-using agents and provides practical insights for designing future systems with stronger temporal reasoning and coordination capabilities.

Problem

• asynchronous tool calling
• multi-task scenarios
• tool response latency
• LLM-based agents
• temporal reasoning

Innovation

• asynchronous tool calling
• multi-task scenarios
• tool response latency
• efficiency-oriented evaluation
• temporal reasoning