HPLT3 Persian Quality Classifier
The repository publishes trained sequence-classification weights and a tokenizer. Its metadata identifies FacebookAI/xlm-roberta-large as the base model, but the README does not explain the training task in detail.
Research release
The trained task
The configuration uses 24 layers, hidden width 1,024 and single-label classification. It contains 8,194 position embeddings. The card does not define the quality labels, training data, thresholds or input procedure, so those settings cannot be recovered from the model name alone.HPLT3 Persian Quality Classifier: model card and trained artifactsHPLT3 Persian Quality Classifier: output labels and configuration
Practical assessment
Practical guidance. Treat this release as a research artifact until the label meanings and input procedure are established. Do not use a confidence score as an educational-quality score merely because another web-filtering model does so.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | quality_classifier_HPLT3_Persian |
| Ancestry | FacebookAI/xlm-roberta-large, as identified by metadata.HPLT3 Persian Quality Classifier: model card and trained artifacts |
| Architecture | XLM-R sequence classifier with 24 layers.HPLT3 Persian Quality Classifier: output labels and configuration |
| License | Apache 2.0.HPLT3 Persian Quality Classifier: model card and trained artifacts |
Related models
- FinerWeb Quality Classifier
TurkuNLP ยท Web quality classification
FinerWeb predicts line-level quality classes for English web text and includes a clean-score calibration component.
Launch date not confirmedDownloadable