What it measures
Generalist web navigation benchmark across 137 real-world websites (Online-Mind2Web evaluates against live websites).
Reading the measurement
The useful outcome is a verified change or deliverable. Inspect the tool interface, initial state, user simulator and policy constraints. An agent can claim completion without reaching the required state, or reach it through an action the protocol forbids.
Published evaluation
Task success evaluated by humans. Automatic evaluator agreement rates are meta-evaluation metrics and are not substituted for task success.
Where this fits in an evaluation suite
A transcript that says “done” is weaker evidence than the verified final state. Track policy violations, recoveries, and invalid attempts alongside success.
Start with a capability hypothesis, then choose a second test with a different failure mode. Keep the checkpoint, reasoning effort, agent tools, sampling policy, and scoring procedure fixed when measuring a training intervention. Record failures and missing outputs in the denominator.
A reproducible comparison
Task success evaluated by humans. Automatic evaluator agreement rates are meta-evaluation metrics and are not substituted for task success.
Publish the exact run configuration, item coverage, score units, and available uncertainty. If a source supplies only an aggregate or a single run, preserve that limit. A new model name or a rounded score alone does not establish a capability improvement.
PRIMARY EVIDENCE & FURTHER READING
- An Illusion of Progress? Assessing the Current State of Web Agents — published result table ↗
All 8 authors
Xue, Tianci, Qi, Weijian, Shi, Tianneng, Song, Chan Hee, Gou, Boyu, Song, Dawn, Sun, Huan, Su, Yu
Reviewed 2026-10-26. Explanation and interpretation by RLTasks. Benchmark ownership remains with its original authors.