Skip to content

Dataset · Congliu

Chinese DeepSeek R1 distillation SFT dataset

Congliu/Chinese-DeepSeek-R1-Distill-data-110k-SFT

110k Chinese examples with R1 answers gathered in the output field for SFT training.

What's inside

A Chinese distillation dataset from the 'full' DeepSeek R1 containing 110k examples. Prepared for SFT: thoughts and answer are combined in the output field, allowing data loading by most SFT frameworks. Includes math as well as a large amount of general data types; language is Chinese.

What you can do with it

  • Fine-tuning a Chinese model on R1 distillations
  • LoRA fine-tuning for math and general tasks
  • Reproducing the R1 distillation effect in Chinese
Tasks
Text generation, Question answering
Languages
Chinese
Size
100K–1M rows
Downloads this month
4.7K
Likes
233

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("Congliu/Chinese-DeepSeek-R1-Distill-data-110k-SFT")