AgentWorld：从多轮交互中训练 LLM 智能体的数据工厂

View PDF HTML (experimental)

Abstract:Training LLMs to act as capable agents requires learning from multi-turn interactions with environments. However, multi-turn interaction trajectories generated by heterogeneous agent runtimes are difficult to transform into structured, high-quality training data that training engines can readily consume. In this paper, we propose AgentWorld, a data factory that bridges heterogeneous agent runtimes and training engines through a Gateway Server and a Data Pool, reframing LLM agent training around a unified data lifecycle from interaction collection to training consumption. The Gateway Server captures multi-turn interactions through an LLM API proxy, while the Data Pool organizes them into standardized step records for curation and training. An interactive dashboard supports trajectory monitoring, step inspection, and data review. In our demonstration, users integrate ready-to-use white-box and black-box agents, trace and curate their interaction data across the pipeline, and extend the factory to new agents and datasets through lightweight configuration. Quantitative experiments across multiple datasets and RL methods demonstrate training support for different agent runtimes. AgentWorld provides a practical foundation for transforming heterogeneous multi-turn experience into training-ready data for LLM agents. Our code is available at this https URL and the demonstration video can be found at link this https URL.

Comments:	Accepted as a demonstration paper at ICDM 2026
Subjects:	Machine Learning (cs.LG); Computation and Language (cs.CL)
Cite as:	arXiv:2606.09138 [cs.LG]
	(or arXiv:2606.09138v2 [cs.LG] for this version)
	https://doi.org/10.48550/arXiv.2606.09138 arXiv-issued DOI via DataCite

Submission history

From: Daoyu Wang [view email]
[v1] Mon, 8 Jun 2026 07:35:18 UTC (1,185 KB)
[v2] Wed, 14 Oct 2026 02:20:50 UTC (787 KB)