This dataset contains the GSM8k dataset formatted for use with open-instruct - specifically reinforcement learning with verifiable rewards.
Part of the Tulu 3 release, for which you can see models here and datasets here.
Each example in the dataset contains the standard instruction-tuning data points as follow:
messages (list): inputs used to prompt the model (after chat template formatting).
ground_truth (str): the… See the full description on the dataset page:
https://huggingface.co/datasets/allenai/RLVR-GSM.
