schwartz-lab-NLP

label-bias

Evaluating Label Bias in LLMs

TOVA

Token Omission Via Attention

fight-bias-with-bias

Code for paper 'Fighting Bias with Bias: Promoting Model Robustness by Amplifying Dataset Biases'

SWEET

Separating Weights for Early-Exit Transformers

papa

Code for the PAPA paper

model_size_and_gender_bias

Code for "Fewer Errors, but More Stereotypes? The Effect of Model Size on Gender Bias" by Yarden Tal, Inbal Magar and Roy Schwartz, 4rd Workshop on Gender Bias in Natural Language, NAACL 2022.

data_contamination