datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
cricketbench
CricketBench — v0.0.5 (preview)
A domain benchmark for cricket reasoning in large language models.
⚠️ Preview release. Contains 56 human-approved questions across 5 dimension(s). Not yet a valid evaluation surface. Public v0.1 release will follow once n ≥ 30 per dimension is reached across all five dimensions.
What's in this release
Dimension
Approved
Status
laws_and_rule_precision
18
Preview release
match_situation_reasoning
9
Preview release… See the full description on the dataset page: https://huggingface.co/datasets/cricketbench/cricketbench.cricket-wiki
cricket-wiki
# data preparation steps
pip install wikiextractor
wget -c https://dumps.wikimedia.org/enwiki/latest/enwiki-latest-pages-articles-multistream.xml.bz2
# extract files using wikiextractor (take a few hours)
python3 -m wikiextractor.WikiExtractor enwiki-latest-pages-articles-multistream.xml.bz2 --json
# get cricket records in a separate file
# take a few minutes
grep -i cricket text/*/* > cricket.jsonl