Tag
This study reveals a 'Smart Pruning Paradox' where activation-aware pruning methods like Wanda preserve perplexity but significantly amplify bias in Large Language Models deployed on edge devices.