{"slug": "weblica-scalable-and-reproducible-training-environments-for-visual-web-agents", "title": "Weblica: Scalable and Reproducible Training Environments for Visual Web Agents", "summary": "Researchers introduced Weblica, a framework for creating scalable and reproducible web environments for training visual web agents. By using HTTP-level caching and LLM-based environment synthesis, they scaled RL training to thousands of diverse environments. Their Weblica-8B model outperformed open-weight baselines across multiple benchmarks and was competitive with API models.", "body_md": "[content type paper](/research/)published July 2026\n\nWeblica: Scalable and Reproducible Training Environments for Visual Web Agents\n\nAuthorsOğuzhan Fatih Kar, Roman Bachmann, Yuanzheng Gong, Anders Boesen Lindbo Larsen, Afshin Dehghan\n\nWeblica: Scalable and Reproducible Training Environments for Visual Web Agents\n\nAuthorsOğuzhan Fatih Kar, Roman Bachmann, Yuanzheng Gong, Anders Boesen Lindbo Larsen, Afshin Dehghan\n\nThe web is complex, open-ended, and constantly changing, making it challenging to scale training data for visual web agents. Existing data collection attempts remain limited to offline trajectories for supervised fine-tuning or a handful of simulated environments for RL training, thus failing to capture web diversity. We propose Weblica (Web Replica), a framework for constructing reproducible and scalable web environments. Our framework leverages 1) HTTP-level caching to capture and replay stable visual states while preserving interactive behavior and 2) LLM-based environment synthesis grounded in real-world websites and core web navigation skills. Using this framework, we scale RL training to thousands of diverse environments and tasks. Our best model, Weblica-8B, outperforms open-weight baselines of similar size across multiple web navigation benchmarks while using fewer inference steps, scales favorably with additional test-time compute, and is competitive with API models.\n\nRephrasing the Web: A Recipe for Compute and Data-Efficient Language Modeling\n\nAugust 6, 2024[research area Methods and Algorithms](/research/?domain=Methods%20and%20Algorithms), [research area Speech and Natural Language Processing](/research/?domain=Speech%20and%20Natural%20Language%20Processing)[conference ACL](/research/?event=ACL)\n\nLarge language models are trained on massive scrapes of the web, which are often unstructured, noisy, and poorly phrased. Current scaling laws show that learning from such data requires an abundance of both compute and data, which grows with the size of the model being trained. This is infeasible both because of the large compute costs and duration associated with pre-training, and the impending scarcity of high-quality data on the web. In this…\n\nRephrasing the Web: A Recipe for Compute and Data-Efficient Language Modeling\n\nMay 7, 2024[research area Methods and Algorithms](/research/?domain=Methods%20and%20Algorithms), [research area Speech and Natural Language Processing](/research/?domain=Speech%20and%20Natural%20Language%20Processing)[Workshop at ICLR](/research/?event=ICLR%20Workshop)\n\nThis paper has been accepted at the Data Problems for Foundation Models workshop at ICLR 2024.\n\nLarge language models are trained on massive scrapes of the web, which are often unstructured, noisy, and poorly phrased. Current scaling laws show that learning from such data requires an abundance of both compute and data, which grows with the size of the model being trained. This is infeasible both because of the large compute costs and duration…", "url": "https://wpnews.pro/news/weblica-scalable-and-reproducible-training-environments-for-visual-web-agents", "canonical_source": "https://machinelearning.apple.com/research/weblica-visual-web-agents", "published_at": "2026-07-07 00:00:00+00:00", "updated_at": "2026-07-07 14:59:30.163731+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "large-language-models", "ai-agents", "computer-vision"], "entities": ["Weblica", "Oğuzhan Fatih Kar", "Roman Bachmann", "Yuanzheng Gong", "Anders Boesen Lindbo Larsen", "Afshin Dehghan"], "alternates": {"html": "https://wpnews.pro/news/weblica-scalable-and-reproducible-training-environments-for-visual-web-agents", "markdown": "https://wpnews.pro/news/weblica-scalable-and-reproducible-training-environments-for-visual-web-agents.md", "text": "https://wpnews.pro/news/weblica-scalable-and-reproducible-training-environments-for-visual-web-agents.txt", "jsonld": "https://wpnews.pro/news/weblica-scalable-and-reproducible-training-environments-for-visual-web-agents.jsonld"}}