WideSearch (reported metric)
Search
Provider-reported WideSearch headline when the release does not identify whether the metric is perfect-table success, row F1 or item F1, or which language split was evaluated. The value is retained as a diagnostic rather than assigned to an invented task denominator or success-rate protocol.
Top models (higher is better)
| Model | Score |
|---|---|
| Atria Dawn Preview | 81.9 |