1 paper · 2 filters
Tianyu Yang, Thy Thy Tran, Iryna Gurevych
Current variational dialog models have employed pre-trained language models (PLMs) to parameterize the likelihood and posterior distributions. However, the Gaussian assumption made…