Can Model Compression Improve NLP Fairness
This work addresses the problem of ensuring fairness in compressed NLP models for safe deployment, though it appears to be an incremental extension of existing regularization concepts to language models.
This paper investigates how model compression techniques affect fairness in NLP models, specifically examining toxicity and bias reduction in GPT2 after applying knowledge distillation and pruning methods.
Model compression techniques are receiving increasing attention; however, the effect of compression on model fairness is still under explored. This is the first paper to examine the effect of distillation and pruning on the toxicity and bias of generative language models. We test Knowledge Distillation and Pruning methods on the GPT2 model and found a consistent pattern of toxicity and bias reduction after model distillation; this result can be potentially interpreted by existing line of research which describes model compression as a regularization technique; our work not only serves as a reference for safe deployment of compressed models, but also extends the discussion of "compression as regularization" into the setting of neural LMs, and hints at the possibility of using compression to develop fairer models.