You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi,
As you provided in the paper, for each vulnerable token you make some samples (perhaps 200) with a moving window. My question is do you prevent the samples from one token to be common in training and test sets? Because there is many common tokens between them, results in optimistic performance.
The text was updated successfully, but these errors were encountered:
Hi,
As you provided in the paper, for each vulnerable token you make some samples (perhaps 200) with a moving window. My question is do you prevent the samples from one token to be common in training and test sets? Because there is many common tokens between them, results in optimistic performance.
The text was updated successfully, but these errors were encountered: