A dataset containing:

'texts': rows of 'hard_text' from the 'train' split of the LabHC/bias_in_bios dataset,

and 'labels' categorizing the instances into 3 classes, based on the fairness of a Bert model fine-tuned on the LabHC/bias_in_bios dataset, where 1 represents the false negative results for the discriminated class(women) of the sensitive attribute(gender), -1 represents the false positive results for the privileged class(men), and 0 shows all the true predictions.