WebArena-Verified

WebArena-Verified tests web agents on realistic, self-hosted websites (such as an online shop or a wiki). The model operates the site like a human user — navigation, forms, shopping carts — and must achieve the stated goal. Tasks with broken success criteria were filtered out in the verified revision.

WebArena-Verified measures how well a model acts as a general-purpose agent. In detail, it checks:

• interacting with realistic web applications
• click paths that reach the concrete goal state
• correct handling of forms, filters and long page lists
• task completion judged against an automatic checker

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better.

What is the WebArena-Verified benchmark?

WebArena-Verified tests web agents on realistic, self-hosted websites (such as an online shop or a wiki). The model operates the site like a human user — navigation, forms, shopping carts — and must achieve the stated goal. Tasks with broken success criteria were filtered out in the verified revision.

What does WebArena-Verified test?

WebArena-Verified mainly measures how well a model acts as a general-purpose agent. Concretely, it checks: interacting with realistic web applications; click paths that reach the concrete goal state; correct handling of forms, filters and long page lists.

How is the WebArena-Verified score calculated?

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better. The declared maximum score of this benchmark is 100.0.

Where do I get WebArena-Verified data and scores?

The official benchmark page/reference is linked on this page. Every score on Open LLM Radar is also available as open data: download the scores JSON (http://open-llm-radar-api:8000/api/v1/benchmarks/webarena-verified/scores) or pull it directly from the Open LLM Radar API with any HTTP client.