1 paper
Junhong Liu, Yuan Zhang, Tao Huang +2
Knowledge distillation (KD) has proven highly effective for compressing large models and enhancing the performance of smaller ones. However, its effectiveness diminishes in cross-m…