BERT Russian toxicity classifier
According to the model card, this BertForSequenceClassification model targets Russian toxic-comment classification with about 178M parameters.
Open Source Model Profile · s-nlp
russian_toxicity_classifier
russian_toxicity_classifier is a 178M-parameter BERT classifier from s-nlp for Russian toxic-comment detection, finetuned from Conversational RuBERT.
- Publisher
- s-nlp
- Task
- text-classification
- Model type
- bert
- License
- openrail++
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
russian_toxicity_classifier is published by s-nlp as a BERT text-classification model for Russian toxic comments. The captured configuration identifies BertForSequenceClassification with model type bert, and Safetensors metadata reports 177,855,490 parameters. According to the model card, it was finetuned from Conversational RuBERT on merged Russian-language toxic-comment datasets.
Recorded capabilities
Merged Russian comment training data
According to the model card, training merged Russian toxic-comment datasets from 2ch.hk and ok.ru with an 80-10-10 train, dev, and test split.
Publisher-reported test metrics
According to the model card, the test set reports 0.97 accuracy with 0.96 macro-average F1 across 26,270 examples.
OpenRAIL++ with usage code
Card data records openrail++ licensing, and the publisher documents Transformers tokenizer and model-loading code.
Use cases in the source record
- Russian-language toxic-comment classification for moderation or filtering workflows using the publisher's documented classifier.
- Benchmarked toxicity-detection experiments that reuse the reported test metrics and the documented Transformers loading example.
Limitations and unknowns
- No context-window value was extracted from this record.
- Reported metrics are publisher test-set claims and were not independently verified by Ethen.
- Provider state is historical snapshot data and should be refreshed before being presented as current.
Source and provenance
Source: s-nlp/russian_toxicity_classifier
Captured: Unknown. Processed: 2026-09-07T19:34:57.200697+00:00.
Bert-based classifier (finetuned from Conversational Rubert ) trained on merge of Russian Language Toxic Comments dataset collected from 2ch.hk and Toxic Russian Comments dataset collected from ok.ru. The datasets were merged, shuffled, and split into train, dev, test splits in 80-10-10 proportion. The metrics obtained from test dataset is as follows precision recall f1-score support 0 0.98 0.99 0.98 21384 1 0.94 0.92 0.93 4886 accuracy 0.97 26270 macro avg 0.96 0.96 0.96 26270 weighted avg 0.97 0.97 0.97 26270 How to use from transformers import BertTokenizer, BertForSequenceClassification # load tokenizer and model weights tokeniz…
F001F002F003F004F005F006F007F009F010F011F012F013F014