Skip to content

Dataset · zai-org

Long-context QA with citations

zai-org/LongCite-45k

44.6k long-context question-answer pairs with sentence-level citations.

What's inside

The dataset contains 44,600 long-context question-answering examples in English and Chinese. Each example includes an instruction, a sentence-split long context, and an answer with sentence-level citations; contexts are up to 128,000 words. The data is designed to train long-context LLMs to generate answers and precise citations in a single output.

What you can do with it

  • Training long-context LLMs with citations
  • SFT for long-document QA
  • Checking citation accuracy
Tasks
Text generation, Question answering
Languages
English, Chinese
Size
10K–100K rows
Downloads this month
1.6K
Likes
78

How to load it

from datasets import load_dataset

ds = load_dataset("zai-org/LongCite-45k")