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

2026-04-20

AI Summary

Training general-purpose agents faces significant challenges due to the scarcity of real-world environments and inadequate mechanisms for continual learning. This work proposes a self-evolving training platform that enables co-evolution between agents and environments through an autonomous environment-task discovery mechanism and a multi-environment reinforcement learning framework. The platform supports dynamic task synthesis driven by capability gaps and integrates a Model Context Protocol toolchain alongside a self-evolving agent arena. Experimental results demonstrate that the trained Agent-World-8B and 14B models substantially outperform strong baselines across 23 benchmarks, confirming the critical role of environmental diversity and iterative self-evolution in enhancing agent intelligence.

Technology Category

• Multiagent Systems: Adversarial Agents
• Machine Learning: Evolutionary Learning
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

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 life-long learning. In this paper, we present \textbf{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-8B and 14B consistently outperforms strong proprietary models and environment scaling baselines. Further analyses reveal scaling trends in relation to environment diversity and self-evolution rounds, offering insights for building general agent intelligence.

Problem

• general agent intelligence
• real-world environment synthesis
• lifelong learning
• agent training
• environment scalability

Innovation

• Agent-World
• self-evolving training
• environment-task synthesis
• multi-environment reinforcement learning
• general agent intelligence