Is the model provided by the author the best model?

TonyNemo / UBAR-MultiWOZ

AAAI 2021: "UBAR: Towards Fully End-to-End Task-Oriented Dialog System with GPT-2"

96 stars 25 forks source link

Is the model provided by the author the best model? #6

Open newcolour1994 opened 3 years ago

newcolour1994 commented 3 years ago

We use the generated BS to query DB results and here are the results in end-to-end setting on WM 2.0 inform 91.5 success 77.4 bleu 17.0 score 101.5

I use the model provided by author, but I can not reproduce the results of end-to-end modeling. The result of my reproduction is 20 points lower than that provided by the author.