---
license: apache-2.0
datasets:
- tay-yozhik/SyntheticTexts
language:
- ru
---
# SyntheticRoBERTa

This is a pre-trained model of type [RoBERTa](https://arxiv.org/abs/1907.11692). 
SyntheticRoBERTa is built on a dataset of texts in Russian, which were generated according to [the rules written in context-free grammar](https://github.com/GorbachevaTaisia/JSGF_generative_grammar).

# Evaluation

This model was evaluated on [RussianSuperGLUE tests](https://russiansuperglue.com/):
| Task | Result | Metrics |
|-------|----------|---------|
| LiDiRus | 0,0 | Matthews Correlation Coefficient |
| RCB | 0,091 / 0,158 | F1 / Accuracy |
| PARus | 0,502 | Accuracy |
| TERRa | 0,487 | Accuracy |
| RUSSE | 0,587 | Accuracy |
| RWSD | 0,331 | Accuracy |