medical dialogue generation is an important yet challenging task. Most
previous works rely on the attention mechanism and large-scale pretrained
language models. However, these methods often fail to acquire pivotal
information from the long dialogue history to yield an accurate and inf