MAG: New Benchmark Unifies Web Agent Actions and Guide Generation

Jul 14, 2026

Researchers introduce MAG, the first benchmark to unify web agent task execution and guide text generation into a single multimodal task. The benchmark uses screenshots with two grounding schemes and provides a complete harness for annotation, training, and evaluation. Frontier models complete fewer than 40% of tasks, and a GRPO training method nearly doubles the success rate of a 9B agent from 6.9% to 13.2%.

Why it matters: MAG provides a unified evaluation framework for web agent actions and guide generation, potentially advancing the development of more capable AI assistants.

Full story at: arXiv AI/ML ↗