BiasGym: A Simple and Generalizable Framework for Analyzing and Removing Biases through Injection
The proposed framework enables consistent bias elicitation for better localization of bias conceptual association in the model space, supports targeted debiasing without degrading performance on downstream tasks, and generalizes to biases unseen during fine-tuning.
Sekh Mainul Islam, Nadav Borenstein, Siddhesh Pawar et al.
· 1 citation