Skip to content

Dataset · zai-org

SFT examples for ultra-long LLM responses

zai-org/LongWriter-6k

6K English and Chinese SFT examples with responses of 2K to 32K words.

What's inside

The dataset contains 6,000 SFT examples for generating ultra-long responses. Output lengths range from 2K to 32K words in English and Chinese. It helps expand the maximum LLM output window to 10K words or more, and the card also lists open models trained on it.

What you can do with it

  • SFT for long-response LLMs
  • Training LongWriter-glm4-9b and similar models
  • Testing generation of 10K+ words
Tasks
Text generation
Languages
English, Chinese
Size
1K–10K rows
Downloads this month
2.3K
Likes
213

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("zai-org/LongWriter-6k")