NEWFairness Pruning: Localizing and Mitigating Demographic Bias in LLM Architectures1. August 2026AI Models, RegulationFairness Pruning localizes demographic bias in GLU-MLP layers by analyzing differential neuron activation and removes it with minimal capability loss. Share on:
