ZhengyanShi/StepGame

Dataset Description We have three files in the dataset (k is the number of maximum hops required to answer the question in the dataset): train.json: The "TrainVersion" is utilised in the baseline models presented in our paper. We use k=1,2,3,4,5 for training without noise. valid.json: The "TrainVersion" is utilised in the baseline models presented in our paper. We use k=1,2,3,4,5 for validation without noise. test.json: The "TrainVersion" is utilised in the baseline models… See the full description on the dataset page: https://huggingface.co/datasets/ZhengyanShi/StepGame.