arXiv 2026 Preprint

Agent-World: Scaling Real-World Environment Synthesis for Evolving General Agent Intelligence

A self-evolving training arena that synthesizes executable environments and verifiable tasks from real-world themes.

Guanting Dong, Junting Lu, Junjie Huang, Wanjun Zhong, Longxiang Liu, Shijue Huang, …, Qinyu Luo et al.

Full author list

Guanting Dong, Junting Lu, Junjie Huang, Wanjun Zhong, Longxiang Liu, Shijue Huang, Zhenyu Li, Yang Zhao, Xiaoshuai Song, Xiaoxi Li, Jiajie Jin, Yutao Zhu, Hanbin Wang, Fangyu Lei, Qinyu Luo, Mingyang Chen, Zehui Chen, Jiazhan Feng, Ji-Rong Wen, Zhicheng Dou

AgentWorld pipeline for environment themes, database mining, tool verification, and task synthesis. — Fig. 2 — Original AgentWorld environment and task discovery pipeline.Original source ↗

01 / The question

How can agent training keep expanding its environments as the agent’s abilities improve?

02 / The approach

Agent-World

Agent-World combines environment and task discovery with multi-environment reinforcement learning. A self-evolving arena identifies capability gaps and creates targeted tasks, allowing the training environments and agent policy to develop together.

Read the full paper

03 / Citation

Cite this work

BibTeX

@misc{dong2026agentworld,
  title = {Agent-World: Scaling Real-World Environment Synthesis for Evolving General Agent Intelligence},
  author = {Dong, Guanting and Lu, Junting and Huang, Junjie and Zhong, Wanjun and Liu, Longxiang and Huang, Shijue and Li, Zhenyu and Zhao, Yang and Song, Xiaoshuai and Li, Xiaoxi and Jin, Jiajie and Zhu, Yutao and Wang, Hanbin and Lei, Fangyu and Luo, Qinyu and Chen, Mingyang and Chen, Zehui and Feng, Jiazhan and Wen, Ji-Rong and Dou, Zhicheng},
  year = {2026},
  eprint = {2604.18292},
  archivePrefix = {arXiv},
  primaryClass = {cs.AI},
  url = {https://arxiv.org/abs/2604.18292}
}

Select the citation to copy it manually.

04 / Versions & records

Publication record

2 Google Scholar records

READ HERE · 48 PAGES

Full paper

Document

Loading document…

Scroll through every page. Select text to copy, or use search to find a passage.

More research

View all

How can agent training keep expanding its environments as the agent’s abilities improve?

Agent-World

Cite this work

Publication record

Full paper

More research

TTExplore↗

UI-TARS-2↗

TTExplore

UI-TARS-2