microsoft/DialogRPT-updown

DialogRPT is a set of dialog response ranking models proposed by Microsoft Research NLP Group trained on 100 + millions of human feedback data . It can be used to improve existing dialog generation model (e.g., DialoGPT) by re-ranking the generated response candidates . The updown score predicts how likely the response is getting upvoted . The task is for the updown task, and other model cards can be found in table below . We considered the following tasks and provided corresponding pretrained models . The tasks are based on human feedback and human-like (human vs fake) human responses . The training and evaluation process is based on training with large-scale human feedback,