WebWorld: A Large-Scale World Model for Web Agent Training
This addresses the need for scalable and safe training environments for web agents, offering a replicable world model with cross-domain applications.
The paper tackles the problem of training web agents by introducing WebWorld, a large-scale open-web simulator trained on over 1 million interactions, which improves Qwen3-14B's performance on WebArena by +9.2% and achieves simulation performance comparable to Gemini-3-Pro.
Web agents require massive trajectories to generalize, yet real-world training is constrained by network latency, rate limits, and safety risks. We introduce \textbf{WebWorld} series, the first open-web simulator trained at scale. While existing simulators are restricted to closed environments with thousands of trajectories, WebWorld leverages a scalable data pipeline to train on 1M+ open-web interactions, supporting reasoning, multi-format data, and long-horizon simulations of 30+ steps. For intrinsic evaluation, we introduce WebWorld-Bench with dual metrics spanning nine dimensions, where WebWorld achieves simulation performance comparable to Gemini-3-Pro. For extrinsic evaluation, Qwen3-14B trained on WebWorld-synthesized trajectories improves by +9.2\% on WebArena, reaching performance comparable to GPT-4o. WebWorld enables effective inference-time search, outperforming GPT-5 as a world model. Beyond web simulation, WebWorld exhibits cross-domain generalization to code, GUI, and game environments, providing a replicable recipe for world model construction.