This is an African American Vernacular English (AAVE) dialect rewritten subset of AsyncHow (rewritten by human annotators). The original data of the subset is obtained by stratified sampling based on complexity.
You can also find a full AsyncHow dataset here, and AAVE-rewritten variations of other popular benchmarks (e.g. GSM8k, HumanEval) here.
If you find our dataset useful, please cite our works below.
@inproceedings{lingraph,
title={Graph-enhanced Large Language Models in Asynchronous… See the full description on the dataset page:
https://huggingface.co/datasets/fangrulin/asynchow_aave.