s1K is a dataset of 1,000 examples of diverse, high-quality & difficult questions with distilled reasoning traces & solutions from Gemini Thining. Refer to the s1 paper for more details.
Usage
pip install -q datasets
from datasets import load_dataset
ds = load_dataset("simplescaling/s1K")["train"]
ds[0]
