AutomationBench-AA - release-reported strict pass rate
Agents
Model-card interpretation of AutomationBench-AA as all end-state assertions passing. The AA feed records partial objective completion; this inconsistent strict-pass claim is diagnostic until the publisher reconciles the metric.
Top models (higher is better)
| Model | Score |
|---|---|
| Muse Wonder Full | 45.4 |
| Muse Wonder | 39.1 |