special token handling

[oathaha]

In the paper you said that 2 special tokens <StartLoc> and <EndLoc> are added in input lines (vulnerable lines). Can you tell me whether these tokens are differently treated or they are treated as normal tokens?

For example, if a model identifies <StartLoc> and <EndLoc> tokens, the model will do something different from what it does to normal token.

[nashid]

@oathaha, after reading the paper, I see that tokens are added as part of the input. The model does not treat any token differently.

[SteveKommrusch]

Correct, the special tokens identifying the 1st buggy line are added before feeding the input to VRepair (such as by identifying risky lines using a tool like Infer). Beyond that, VRepair learns based on training examples that these tokens identify the 1st buggy line and it learns to make the first proposed modification within this line. No further special processing needs to be done. VRepair may learn that further lines must be modified based on training experience.

Regards, Steve

On Mon, Apr 11, 2022 at 11:41 AM nashid @.***> wrote:

@oathaha https://github.com/oathaha, after reading the paper, I see that tokens are added as part of the input. The model does not treat any token differently.

— Reply to this email directly, view it on GitHub https://github.com/SteveKommrusch/VRepair/issues/7#issuecomment-1095343703, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGFBJG7TPPDI4FD344G63I3VERP3VANCNFSM5TDBVLPA . You are receiving this because you are subscribed to this thread.Message ID: @.***>