uclnlp / EMAT

Efficient Memory-Augmented Transformers
35 stars 4 forks source link

Replicating baselines form Paper #3

Open DanielRoeder1 opened 1 year ago

DanielRoeder1 commented 1 year ago

Hi,

I am really interested in your paper and I am trying to replicate some baselines of yours to compare to. In the paper you state parametric models as one of the baselines. Am I correct in understanding that parametric models receive the questions as input and are tasked with generating the output? (i.e. no knowledge in form of context passages is fed to the model)

If this is the case I am struggling to reach the same kind of EM / F1 on these baselines (specifically t5-base). Any advise is welcome1