Views
No views yet
all), using the unredacted doxxing text for the items that are
redacted in the public dataset.| test set | n | macro F1 | TFGBV F1 | accuracy |
|---|---|---|---|---|
| ALTO gold test (unredacted) | 349 | 0.7794 | 0.7440 | 0.7851 |
| bench test | 1878 | 0.7403 | 0.5330 | 0.9058 |
BinaryClassifier module
defined in the repo notebook. See that notebook for load_best_model and predict.