Attention Pruning: Automated Fairness Repair of Language Models via Surrogate Simulated Annealing

AAML
Re-assign community