Fetching the paper…

AD-KD: Attribution-Driven Knowledge Distillation for Language Model Compression · Around