{"slug": "web-agents-world-models-learn-to-predict-not-to-choose-a-rebuilt-training-lifts", "title": "Web agents' world models learn to predict, not to choose: a rebuilt training objective lifts task success", "summary": "A paper posted 2 September by a team including Kelvin Li, Leonid Karlinsky, Rogerio Feris, Trevor Darrell and Roei Herzig argues that LLM web agent pipelines train world models with supervised next-state prediction to reproduce pages faithfully, while the process reward model ranker actually needs predictions that discriminate between alternative actions. Their predicted-state matching objective trains the world model to distinguish the true resulting state from states other actions would reach, using a branching dataset built from WebArena Go-Browse trajectories. The method beats supervised next-state prediction on their held-out benchmark, improves process-reward action ranking on WebPRMBench, and raises end-to-end task success on WebArena-Lite, drawing researcher endorsements on SciRate within hours of posting.", "body_md": "LLM web agents increasingly pick actions by sampling candidate moves, predicting the resulting page states with a world model, and letting a process reward model rank the options. A paper posted 2 September by a team including Kelvin Li, Leonid Karlinsky, Rogerio Feris, Trevor Darrell and Roei Herzig argues that pipeline hides a flaw: world models are trained with supervised next-state prediction to reproduce pages faithfully, but the ranker actually needs predictions that discriminate between alternative actions. Their predicted-state matching objective trains the world model to tell the true resulting state apart from states other actions would reach, using a branching dataset built from WebArena Go-Browse trajectories. It beats supervised next-state prediction on their held-out benchmark, improves process-reward action ranking on WebPRMBench, and raises end-to-end task success on WebArena-Lite. The paper drew researcher endorsements on SciRate within hours of posting.", "url": "https://wpnews.pro/news/web-agents-world-models-learn-to-predict-not-to-choose-a-rebuilt-training-lifts", "canonical_source": "https://provenbrief.com/story/web-agents-world-models-learn-to-predict-not-to-choose-a-rebuilt-training-object", "published_at": "2026-09-03 16:46:57+00:00", "updated_at": "2026-09-10 21:42:49.505170+00:00", "lang": "en", "topics": ["ai-agents", "ai-research", "large-language-models", "ai-tools"], "entities": ["Kelvin Li", "Leonid Karlinsky", "Rogerio Feris", "Trevor Darrell", "Roei Herzig", "WebArena", "WebPRMBench", "SciRate"], "alternates": {"html": "https://wpnews.pro/news/web-agents-world-models-learn-to-predict-not-to-choose-a-rebuilt-training-lifts", "markdown": "https://wpnews.pro/news/web-agents-world-models-learn-to-predict-not-to-choose-a-rebuilt-training-lifts.md", "text": "https://wpnews.pro/news/web-agents-world-models-learn-to-predict-not-to-choose-a-rebuilt-training-lifts.txt", "jsonld": "https://wpnews.pro/news/web-agents-world-models-learn-to-predict-not-to-choose-a-rebuilt-training-lifts.jsonld"}}