The data in the paper cannot be reproduced #5
Description
I would like to express my admiration for your research, as it holds great significance in our field.
However, while attempting to replicate your code, I have encountered significant discrepancies in the data. My experimental results differ greatly from those presented in your paper. I have thoroughly reviewed the code and experimental settings, but have been unable to resolve this issue. Therefore, I am reaching out to seek your guidance and advice on potential solutions.
Firstly, I would like to confirm the usage of the dataset. I have employed the dataset mentioned in your paper and followed the preprocessing steps as described. However, I have observed substantial differences in my results compared to yours, using the same dataset.
This is the parsing result obtained using the trained data provided in the original code.
When I executed the flow from freshshot_sampling -> train -> benchmark in README, I found that the GA and PA in most of the datasets have decreased to a great extent, and the execution results are shown below.
I am curious to know if there are different versions of the dataset or any additional preprocessing steps that may account for these discrepancies.
Thank you once again for your contributions and valuable time. I look forward to your response.