r1-zero
Datasets
All datasets matching “r1-zero”ru-thinking-reasoning-r1-v2-dedupedru-thinking-reasoning-r1Combined dataset of mostly Russian thinking/reasoning/reflection dialogs in form of conversation suitable for LLM fine-tuning scenarios. All responses are mapped to same format.
The format of reasoning in most cases is:
<think>
Reasoning...
</think>
Response
For reflection dataset - there can be also <reflection> tags inside <think>.
Common system prompt for think:
Ты полезный ассистент. Отвечай на вопросы, сохраняя следующую структуру: <think> Твои мысли и рассуждения </think>
Твой конечный… See the full description on the dataset page: https://huggingface.co/datasets/ZeroAgency/ru-thinking-reasoning-r1.ru-thinking-reasoning-r1-v2This is a modified version of ZeroAgency/ru-thinking-reasoning-r1 with addition of Egor-AI/CoT-XLang dataset.
Combined dataset of mostly Russian thinking/reasoning/reflection dialogs in form of conversation suitable for LLM fine-tuning scenarios. All responses are mapped to same format.
The format of reasoning in most cases is:
<think>
Reasoning...
</think>
Response
For reflection dataset - there can be also <reflection> tags inside <think>.
Common system prompt for think:
Ты полезный… See the full description on the dataset page: https://huggingface.co/datasets/ZeroAgency/ru-thinking-reasoning-r1-v2.Table-R1-Zero-Datasettrn-r1-zero-tagsru-thinking-reasoning-r1-deduped