Leaderboard
Run list
Every run on every benchmark, newest first. Filter by provider, benchmark, text or date; select a run for its configuration and the model's other runs on that benchmark.
Provider
Loading runs…Every stored run: legacy configurations and runs without a score are included, so a model can appear here that the leaderboard leaves out.
| Each benchmark's own metric (F1, CER, fuzzy match) scaled to 0–100, higher is better. |
|---|
Downloadevery run (CSV)every run with test configurations (JSON)complete dataset (Zenodo)