Analyzing the Forgetting Problem in the Pretrain-Finetuning of Dialogue Response Models

In this work, we study how the large-scale pretrain-finetune framework changes the behavior of a neural language generator. We focus on the transformer encoder-decoder model for the open-domain dialogue response generation task. We find that after standard fine-tuning, the model \textit{forgets} important language generation skills acquired during large-scale pre-training. We demonstrate the forgetting phenomenon through a set of detailed behavior analysis from the perspectives of context sensitivity, knowledge transfer, and function space projection. Adopting the concept of data mixing, we propose an intuitive fine-tuning strategy named "mix-review". We find that mix-review effectively regularize the fine-tuning process, and the forgetting problem is largely alleviated. Finally, we discuss interesting behavior of the resulting dialogue model and its implications.