AD-KD: Attribution-Driven Knowledge Distillation for Language Model Compression
This work addresses model compression for language models, which is an incremental improvement in efficiency for NLP applications.
The paper tackles the problem of compressing pre-trained language models by addressing limitations in existing knowledge distillation methods, such as ignoring underlying reasoning and data-specific knowledge, and introduces an attribution-driven approach that transfers token-level rationale using Integrated Gradients, achieving superior performance on the GLUE benchmark with BERT.
Knowledge distillation has attracted a great deal of interest recently to compress pre-trained language models. However, existing knowledge distillation methods suffer from two limitations. First, the student model simply imitates the teacher's behavior while ignoring the underlying reasoning. Second, these methods usually focus on the transfer of sophisticated model-specific knowledge but overlook data-specific knowledge. In this paper, we present a novel attribution-driven knowledge distillation approach, which explores the token-level rationale behind the teacher model based on Integrated Gradients (IG) and transfers attribution knowledge to the student model. To enhance the knowledge transfer of model reasoning and generalization, we further explore multi-view attribution distillation on all potential decisions of the teacher. Comprehensive experiments are conducted with BERT on the GLUE benchmark. The experimental results demonstrate the superior performance of our approach to several state-of-the-art methods.