1 paper · 1 filter
Jiacheng Lyu, Bihua Bao, Shiyun Yan
Massive datasets often contain redundancy that inflates computational costs without improving generalization. Existing data reduction methods are typically task-agnostic, discardin…