First of all, I have really appreciated this work.
Second, I was wondering whether the code can be found here (https://github.com/awsm-research/VulRepair/tree/main/M8_VRepair_subword) is a re-implementation of the VRepair approach, with the only difference being the data on which the model is trained on. After quickly skimming through the code, some doubts arose.
Moreover, if we focus on Fig 2. (the one in the paper), It's unclear whether CodeT5, CodeBERT, and VRepair have been trained and tested on the same data. Could you please set the record straight on my doubts?
Hi,
First of all, I have really appreciated this work.
Second, I was wondering whether the code can be found here (https://github.com/awsm-research/VulRepair/tree/main/M8_VRepair_subword) is a re-implementation of the VRepair approach, with the only difference being the data on which the model is trained on. After quickly skimming through the code, some doubts arose.
Moreover, if we focus on Fig 2. (the one in the paper), It's unclear whether CodeT5, CodeBERT, and VRepair have been trained and tested on the same data. Could you please set the record straight on my doubts?
Many thanks for your help.