I used your dataset to train baseline models. It gave less accuracy compare to the accuracy mentioned in the papers.