Skip to content

Dataset · allenai

Extreme summaries (TLDR) of scientific articles

allenai/scitldr

5.4 thousand concise summaries for 3.2 thousand scientific papers from AllenAI.

What's inside

Developed by the AllenAI institute, this English-language dataset features 5.4 thousand text summaries (TLDR) covering 3.2 thousand research articles. It combines brief abstracts written directly by the paper authors with summaries produced by domain experts following a specialized annotation protocol.

What you can do with it

  • Training models for extreme one-to-two sentence summarization of scientific texts
  • Evaluating the quality of generated titles and short abstracts for research papers
  • Testing LLM capabilities in extracting the core essence of academic publications
Tasks
Summarization
Languages
English
Size
1K–10K rows
Downloads this month
1.2K
Likes
36

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("allenai/scitldr")