Weblica: a scalable and reproducible training environment for visual web agents
Original title:Weblica: Scalable and Reproducible Training Environments for Visual Web Agents
AI summary
Apple researchers proposed the Weblica framework, which uses HTTP-level caching to preserve stable visual states of web pages while retaining interactive behavior, and combines large language models to synthesize environments based on real websites and core navigation skills, building a reproducible and scalable training environment. The framework scales reinforcement learning training to thousands of diverse environments and tasks. The best model, Weblica-8B, outperforms open-source models of comparable scale on multiple web navigation benchmarks with fewer reasoning steps, scales well with test-time compute, and performs comparably to API models.
Key Takeaways
- Apple researchers proposed the Weblica framework,which uses HTTP-level caching to preserve stable…
- The framework scales reinforcement learning training to thousands of diverse environments and tasks.
- The best model,Weblica-8B,outperforms open-source models of comparable scale on multiple web…
Quick Facts
- Published
- Updated
- Category
- Papers & Research
- AI score
- 56
Apple 的研究把 web agent 训练从零散数据中解放出来,可复现环境是规模化 RL 的关键一步,做 web 自动化的值得关注。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Weblica: a scalable and reproducible training environment for visual web agents. July 7, 2026. https://new.nat6.net/en/news/cmrasd6r300i0ihog8xxwezye/