T-Wix: Russian Supervised Fine-Tuning Dataset for Language Models

T-Wix is a Russian supervised fine-tuning dataset built entirely from publicly available data. The dataset is divided into two subsets, or 'sticks', and is intended for research and development purposes. It was created by the author 't-tech' and was last updated on March 31, 2026.

Use Cases

• Instruction-tuning language models based on the Russian SFT data.
• Evaluating model performance on Russian-language tasks based on the described SFT mixture.
• Researching biases and artifacts in Russian-language AI training data.
• Developing conversational AI assistants for Russian-speaking users.

Strengths

• Built entirely from publicly available data, ensuring transparency.
• Specifically designed for supervised fine-tuning of language models.
• Last updated on March 31, 2026, indicating recent maintenance.

Limitations

• The dataset may contain noise, biases, or artifacts requiring careful inspection.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file formats are unknown, which may limit suitability assessment.

Provenance

Source

Publicly available data, aggregated by author 't-tech'.

Collection Method

Likely collected and curated from various public Russian-language sources.

Freshness

Last updated 2026-03-31 13:18:42.

Geography

Primarily Russian-language content; specific geographic coverage is unknown.