how to generate poison data

[bingtong109]

Hello author~
First of all, thank you for your excellent work on this project. I have a question regarding the poisoned datasets you provided. I trained the poison model as per the instructions and then used the "generate_data.py" script to produce poisoned data. However, the data I generated appears different from the poisoned datasets you've provided. Most importantly, the poisoned data from your dataset leads to significantly better performance when training the recommendation model compared to the data I generated myself.

Could you please explain how your poisoned datasets were generated? I am particularly interested in any specific techniques or parameters that might help me replicate the quality of your datasets.

Thank you for your assistance!