datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
gsm-plus-gt-corrected
GSM-Plus (GT-Corrected) in GSM8K format
GSM-Plus perturbations of GSM8K test with opus-corrected ground-truth labels, formatted to match the openai/gsm8k schema (question / answer ending with #### N). Built to give cleaner numerical-reasoning benchmarks and SFT data: the GSM-Plus release was found to mislabel ~9% of rows in the perturbations we ran, and this dataset surfaces opus's corrections.
Configs
Both splits are test.
Config
Rows
Perturbation… See the full description on the dataset page: https://huggingface.co/datasets/onnookk/gsm-plus-gt-corrected.open-orca-slimorca-deduped-cleaned-corrected-for-pascal-txtThis is a modified version of the slimorca-deduped-cleaned-corrected dataset.
It contains English only characters.
Open Orca Slim for Pascal Developers is a subset of the original Open Orca dataset .
Open Orca Slim for Pascal Developers dataset was created with:
from datasets import load_dataset
# Coded by Gemini
def biggest_char_code(input_string):
"""
Returns the largest character code in a string.
"""
if not input_string:
return None # Handle empty string case
largest_code… See the full description on the dataset page: https://huggingface.co/datasets/schuler/open-orca-slimorca-deduped-cleaned-corrected-for-pascal-txt.