Why does Knowledge Distillation work? Rethink its attention and fidelity mechanism

Does Knowledge Distillation (KD) really work? Conventional wisdom viewed it as a knowledge transfer procedure where a perfect mimicry of the student to its teacher is desired. However, paradoxical studies indicate that closely replicating the teacher's behavior does not consistently improve student generalization, posing questions on its possible causes. Confronted with this gap, we hypothesize that diverse attentions in teachers contribute to better student generalization at the expense of reduced fidelity in ensemble KD setups. Focusing on supervised image classification task, by increasing data augmentation strengths, our key findings reveal a decrease in the Intersection over Union (IoU) of attentions between teacher models, leading to reduced student overfitting and decreased fidelity. We propose this low-fidelity phenomenon as an underlying characteristic rather than a pathology when training KD. This suggests that stronger data augmentation fosters a broader perspective provided by the divergent teacher ensemble and lower student-teacher mutual information, benefiting generalization performance. We further demonstrate that even optimization towards logits-matching between teachers and student can hardly mitigate this low-fidelity effect. These insights clarify the mechanism on low-fidelity phenomenon in KD. Thus, we offer new perspectives on optimizing student model performance, by emphasizing increased diversity in teacher attentions and reduced mimicry behavior between teachers and student. Codes are available at https://github.com/zisci2/RethinkKD