Paper 2601.08225
User-Oriented Multi-Turn Dialogue Generation with Tool Use at scale
- Published
- Jan 2026
- Research lab
- Independent
- Citations
- 1
- GitHub
- Not linked
01 In brief
Summary
This paper introduces a user-oriented simulation framework for generating multi-turn tool-use dialogue data at scale, addressing limitations of static toolsets and single-shot trajectories.
The authors first developed a task-oriented pipeline using an LRM-based simulator to dynamically synthesize tools and tasks, but found it produced minimal-interaction, task-solving trajectories.
They then shifted to a user-oriented paradigm that decouples task generation from a dedicated user simulator, which mimics human behavior by making incremental requests and providing turn-by-turn feedback, resulting in more authentic, extended dialogues.
The pipeline is modular and plug-and-play, enabling generation from any state and producing high-density trajectories with multiple task completions.
They also introduced an executable SQL-backed tool synthesis grounded in real-world database schemas (e.g., Spider) to ensure verifiable tool outputs.
Experiments fine-tuning Qwen3 models on their data showed consistent improvements on BFCL and τ2 benchmarks, particularly in long-horizon and stateful domains like Telecom, with higher Pass^k consistency.
Ablations confirmed that user-oriented simulation with tool execution yields the strongest performance.
The work highlights the importance of user-oriented interaction modeling and execution-grounded supervision for training robust agentic reasoning models.
02 From the paper
Abstract
The recent paradigm shift toward large reasoning models (LRMs) as autonomous agents has intensified the demand for sophisticated, multi-turn tool-use capabilities. Yet, existing datasets and data-generation approaches are limited by static, predefined toolsets that cannot scale to the complexity of open-ended human-agent collaboration. To address this, we initially developed a framework for automated task-oriented multi-turn dialogue generation at scale, utilizing an LRM-based simulator to dynamically generate high-value, domain-specific tools to solve specified tasks. However, we observe that a purely task-oriented design often results in "solely task-solving" trajectories, where the agent completes the objective with minimal interaction, failing to generate the high turn-count conversations seen in realistic scenarios. To bridge this gap, we shift toward a user-oriented simulation paradigm. By decoupling task generation from a dedicated user simulator that mimics human behavioral rules - such as incremental request-making and turn-by-turn feedback - we facilitate more authentic, extended multi-turn dialogues that reflect the iterative nature of real-world problem solving. Our generation pipeline operates as a versatile, plug-and-play module capable of initiating generation from any state, ensuring high scalability in producing extended tool-use data. Furthermore, by facilitating multiple task completions within a single trajectory, it yields a high-density dataset that reflects the multifaceted demands of real-world human-agent interaction.