Qwen-AgentWorld: Language World Models for General Agents

2026-06-23

AI Summary

This work proposes Qwen-AgentWorld, the first language-based world model capable of simulating multi-domain agent environments to enhance general-purpose agents’ reasoning and planning capabilities. Trained through a three-stage pipeline comprising causal pretraining (CPT), supervised fine-tuning (SFT), and reinforcement learning with a hybrid scoring rule, the model is optimized on over 10 million real-world interaction trajectories and demonstrates long-horizon environmental dynamics prediction across seven domains. The study introduces a novel dual-paradigm architecture that decouples the simulator from a unified foundation model and establishes AgentWorldBench, a comprehensive evaluation framework. Experimental results show that Qwen-AgentWorld, when used as a simulator, effectively expands training environments, and as a foundation model, significantly boosts downstream task performance, outperforming existing methods on multiple state-of-the-art benchmarks.

Abstract

A world model predicts environment dynamics based on current observations and actions, serving as a core cognitive mechanism for reasoning and planning. In this work, we investigate how world modeling based on language models can further push the boundaries of general agents. (i) We first focus on building foundation models for agentic environment simulation. We introduce Qwen-AgentWorld-35B-A3B and Qwen-AgentWorld-397B-A17B, the first language world models capable of simulating agentic environments covering 7 domains via long chain-of-thought reasoning. Leveraging more than 10M environment interaction trajectories of 7 domains in real-world environments, we develop Qwen-AgentWorld through a three-stage training pipeline: CPT injects general-purpose world modeling capabilities from the state transition dynamics and augmented professional corpora, SFT activates next-state-prediction reasoning, and RL sharpens simulation fidelity through a tailored framework with hybrid rubric-and-rule rewards. To evaluate language world models, we present AgentWorldBench, a comprehensive benchmark constructed from real-world interactions of 5 frontier models on 9 established benchmarks. Empirical results demonstrate that Qwen-AgentWorld significantly outperforms existing frontier models. (ii) Beyond foundation models, we further investigate two complementary paradigms through which world modeling enhances general agents. First, as a decoupled environment simulator, Qwen-AgentWorld supports scalable and controllable simulation of thousands of real-world environments for agentic RL, yielding gains that surpass real-environment training alone. Second, as a unified agent foundation model, world-model training acts as a highly effective warm-up that improves downstream performance across 7 agentic benchmarks. Code: https://github.com/QwenLM/Qwen-AgentWorld

Problem

• world model
• language models
• general agents
• environment simulation
• agentic reasoning

Innovation

• language world models
• agentic simulation
• chain-of-thought reasoning
• three-stage training pipeline
• AgentWorldBench