top of page

ZeroSCROLLS: Zero-Shot CompaRison Over Long Language Sequences

What is Zero-SCROLLS?

ZeroSCROLLS is a suite of datasets that require synthesizing information over long texts. The benchmark includes ten natural language tasks across multiple domains, including summarization, question answering, aggregated sentiment classification and information reordering.

Citing ZeroSCROLLS

Please use the following bibliography to cite ZeroSCROLLS:

    title = "{Z}ero{SCROLLS}: A Zero-Shot Benchmark for Long Text Understanding",
    author = "Shaham, Uri  and
      Ivgi, Maor  and
      Efrat, Avia  and
      Berant, Jonathan  and
      Levy, Omer",
    editor = "Bouamor, Houda  and
      Pino, Juan  and
      Bali, Kalika",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
    month = dec,
    year = "2023",
    address = "Singapore",
    publisher = "Association for Computational Linguistics",
    url = "",
    doi = "10.18653/v1/2023.findings-emnlp.536",
    pages = "7977--7989"

When citing ZeroSCROLLS, please make sure to cite all of the original dataset papers. [bibtex]

Contact Us

Tel-Aviv University NLP lab
bottom of page