TensorX
返回文献探索

Paper · arXiv 2602.14721

WebWorld: A Large-Scale World Model for Web Agent Training

Zikai Xiao, Jianhong Tu, Chuhang Zou, Yuxin Zuo, Zhi Li, Peng Wang, Bowen Yu, Fei Huang, Junyang Lin, Zuozhu Liu

20 upvotesFebruary 16, 2026arXiv 预印本
AI 摘要

WebWorld is an open-web simulator trained on over one million interactions that supports long-horizon reasoning and multi-format data, achieving performance comparable to advanced models like Gemini-3-Pro and GPT-4o.

web simulatoropen-web environmentintrinsic evaluationextrinsic evaluationworld modelinference-time searchcross-domain generalization

Abstract

Web agents require massive trajectories to generalize, yet real-world training is constrained by network latency, rate limits, and safety risks. We introduce 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.

北京市昌平区探索星信息技术及软件开发工作室

京ICP备2026059466号
WebWorld: A Large-Scale World Model for Web Agent Training | TensorX