Agent-World

Large language models are increasingly expected to serve as general-purpose agents that interact with external, stateful tool environments. The Model Context Protocol (MCP) and broader agent skills offer a unified interface for connecting agents with scalable real-world services, but training robust agents remains limited by the lack of realistic environments and principled mechanisms for lifelong learning.

In this paper, we present Agent-World, a self-evolving training arena for advancing general agent intelligence through scalable environments. Agent-World has two main components: (1) Agentic Environment-Task Discovery, which autonomously explores topic-aligned databases and executable tool ecosystems from thousands of real-world environment themes and synthesizes verifiable tasks with controllable difficulty; and (2) Continuous Self-Evolving Agent Training, which combines multi-environment reinforcement learning with a self-evolving agent arena that automatically identifies capability gaps through dynamic task synthesis and drives targeted learning, enabling the co-evolution of agent policies and environments.

Across 23 challenging agent benchmarks, Agent-World consistently outperforms strong proprietary models and environment scaling baselines. Further analyses reveal scaling trends with environment diversity and self-evolution rounds, offering insights for building general agent intelligence.

As the capability frontier of large language models continues to expand, expectations are shifting from chat-oriented text generation toward general-purpose agent assistants. Ideally, such agents should seamlessly integrate real-world interaction with verbal reasoning, and continuously learn from experience to improve themselves. Realizing these agentic capabilities requires training LLMs in dynamic environments equipped with executable tools, forming a "Generation–Execution–Feedback" interaction loop.

With the rise of agentic reinforcement learning (Agent RL), several agent systems built on static tool environments have demonstrated strong practical value. However, open-world tool environments are inherently compositional and stateful. For instance, in a flight-booking workflow, an agent should follow a valid action order (check inventory → execute booking → update the calendar), while each action also modifies the underlying environment state. Prior work centered on stateless or single-tool settings is insufficient for realistic applications.

Two key bottlenecks remain unresolved:

We propose Agent-World, a general-purpose agent training arena that unifies scalable environment synthesis with continuous self-evolving training. Agent-World follows a two-stage design that forms a closed-loop training process.

Key Contributions

• We introduce Agent-World, a general-purpose agent training arena that unifies scalable environment synthesis with a continuous self-evolving training mechanism, forming a co-evolution loop between agent policies and environments.
• We propose Agentic Environment-Task Discovery, which mines realistic executable environments from real-world environment themes and synthesizes diverse verifiable tasks with controllable difficulty.
• We propose Continuous Self-Evolving Agent Training, which integrates multi-environment agentic RL with a self-evolving arena to automatically diagnose agent weaknesses and drive targeted learning in a closed training loop.
• Experiments across 23 challenging agent benchmarks demonstrate the superior performance of Agent-World. Further analysis reveals scaling relationships among environment diversity, evolution rounds, and agent performance.

Agent-World contains two tightly coupled components that form a closed loop: scalable environments support agent training, while training-time diagnosis feeds back into the next round of environment-task construction.

1 Agentic Environment-Task Discovery

2 Continuous Self-Evolving Agent Training

We evaluate Agent-World on 23 benchmarks spanning agentic tool use, advanced AI assistant, software engineering, deep research, and general reasoning, using Qwen3-8B/14B backbones trained with GRPO.

Main Results on Agentic Tool-Use Benchmarks

We report accuracy (%) across three benchmark suites: MCP-Mark, BFCL V4, and τ²-Bench.

Key Findings

Generalization on Advanced AI Assistant Benchmarks

Scaling Analysis of Training Environments

We progressively increase the number of training environments from 0 to 2000. Performance improves consistently across all domains as the environment scale grows. Averaged over four domains, the score rises from 18.4% to 38.5% (+20.1 points), more than doubling the initial level. The gains are particularly pronounced on interaction-intensive tasks.

Analysis of Continuous Self-Evolution

To validate Continuous Self-Evolving Agent Training, we run the same two-round self-evolving arena loop from two different starting points: Agent-World-14B and EnvScaler-8B. Results show monotonic gains on all three evaluation suites for both models:

The largest gains across two rounds appear on MCP-Mark for both models: +8.6 for Agent-World and +5.6 for EnvScaler. This setting requires stronger state tracking and more reliable interaction with realistic MCP server environments. Importantly, EnvScaler-8B also improves, indicating that the loop not only benefits our base model but also yields sustained gains for other environment-scaling baselines without relying on Agent-World initialization.

Training Dynamics

We presented Agent-World, a self-evolving training arena for general-purpose agents in realistic tool environments. Agent-World unifies two tightly coupled components:

Agentic Environment-Task Discovery mines topic-aligned real-world databases and executable toolsets from large-scale themes and synthesizes verifiable tasks with controllable difficulty.

Continuous Self-Evolving Agent Training combines multi-environment reinforcement learning with an agentic diagnostic arena to identify capability gaps and drive targeted iterative data expansion.

Experiments across 23 challenging benchmarks demonstrate that Agent-World consistently improves performance over strong baselines. Further analyses reveal clear scaling trends with respect to environment diversity, evolution rounds, and task difficulty, suggesting that scalable realistic environments are not only useful data sources, but also critical infrastructure for advancing general agent capabilities.

@article{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},
  journal = {arXiv preprint},
  year    = {2026}
}

We greatly thank Yujia Qin² and Guang Shi² for supporting this work and providing valuable suggestions. We also thank Yifei Chen¹ for valuable discussions.