T1: A Tool-Oriented Conversational Dataset for Multi-Turn Agentic Planning

2025-05-22

AI Summary

Modeling cross-tool dependencies and enabling dynamic planning across multi-turn dialogues constitute a core challenge in agent reasoning. Methodologically, we introduce the first tool-dependency-aware dialogue dataset, T1, covering nine single- and multi-domain task compositions and systematically characterizing explicit inter-API dependencies for the first time. Our approach integrates a dual-timescale caching mechanism with a dynamic replanning strategy—supporting adaptive decisions to reuse or recompute prior steps—alongside tool-augmented dialogue construction, multi-domain task orchestration, and cache-aware inference. Experiments demonstrate that the T1-Agent, trained on T1, significantly improves completion rates and reasoning consistency on multi-step dependent tasks. Concurrently, T1 establishes a new open-source benchmark for rigorously evaluating large language models’ planning capabilities.

Technology Category

• Multiagent Systems: Multiagent Planning
• Planning, Routing, and Scheduling: Planning with Language Models
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Large Language Models (LLMs) have demonstrated impressive capabilities as intelligent agents capable of solving complex problems. However, effective planning in scenarios involving dependencies between API or tool calls-particularly in multi-turn conversations-remains a significant challenge. To address this, we introduce T1, a tool-augmented, multi-domain, multi-turn conversational dataset specifically designed to capture and manage inter-tool dependencies across diverse domains. T1 enables rigorous evaluation of agents' ability to coordinate tool use across nine distinct domains (4 single domain and 5 multi-domain) with the help of an integrated caching mechanism for both short- and long-term memory, while supporting dynamic replanning-such as deciding whether to recompute or reuse cached results. Beyond facilitating research on tool use and planning, T1 also serves as a benchmark for evaluating the performance of open-source language models. We present results powered by T1-Agent, highlighting their ability to plan and reason in complex, tool-dependent scenarios.

Problem

• Addressing multi-turn planning challenges with tool dependencies
• Evaluating agent coordination across diverse tool domains
• Benchmarking open-source models in tool-dependent scenarios

Innovation

• Tool-augmented multi-domain conversational dataset
• Integrated caching for short- and long-term memory
• Dynamic replanning with recompute or reuse decisions