构建首个大规模开放网页模拟器，支持百万级真实交互训练。
WebWorld: A Large-Scale World Model for Web Agent Training
- 基于可扩展数据流训练100万+真实网页交互，支持长序列推理。
- 在WebArena上提升9.2%性能，逼近GPT-4o水平。
- 适用于网页、代码、图形界面等多场景，可复现的模型构建方案。
Web代理需要大量轨迹才能实现泛化，但真实训练受限于网络延迟、频率限制和安全风险。我们提出首个大规模开源网页模拟器WebWorld系列。与仅支持数千条轨迹的封闭环境模拟器不同，WebWorld通过可扩展数据管道训练超过100万次开放网页交互，支持推理、多格式数据和30步以上的长时序模拟。内在评估引入WebWorld-Bench，涵盖九个维度的双指标体系，其仿真性能接近Gemini-3-Pro。外在评估显示，基于WebWorld合成轨迹训练的Qwen3-14B在WebArena上提升9.2%，表现媲美GPT-4o。WebWorld还支持高效的推理时搜索，优于GPT-5作为世界模型。此外，其跨领域泛化能力覆盖代码、GUI和游戏环境，提供可复现的世界模型构建范式。
原文摘要 · Abstract (English)
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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。