README.md

metadata

license: apache-2.0

LoCoV1 Queries

The queries for the LoCoV1 dataset from the paper, "Benchmarking and Building Long-Context Retrieval Models with LoCo and M2-BERT"

How to Use

To load the dataset, use the following command:

from datasets import load_dataset
dataset = load_dataset("hazyresearch/LoCoV1-Queries")

To load a specific subset, such as SummScreenFD, use the following command:

from datasets import load_dataset
dataset = load_dataset("hazyresearch/LoCoV1-Queries")
def filter_condition(example):
    return example['dataset'] == 'summ_screen_fd'
filtered_dataset = dataset.filter(filter_condition)

The datasets available are:

{'summ_screen_fd', 'gov_report', 'qmsum', 'qasper_title', 'qasper_abstract', '2wikimqa', 'multifieldqa',
 'passage_retrieval', 'courtlistener_HTML', 'courtlistener_Plain_Text', 'legal_case_reports', 'stackoverflow'}

For documents, go to hazyresearch/LoCoV1-Documents.