Dear Author,
Your work is very interesting and inspiring.
I would like some clarification on the scenario example you have used for the construction of the prompt template for RedCode-Gen.
Specifically, I am wondering whether the scenario example is just a text description of the potential scenarios specific to the category or a code implementation demonstration.
My confusion came from the following description:
"These prompts are generated with human-in-the-loop AI-assisted generation, where we use GPT-4 to generate candidate prompts using real examples of malware and human-written descriptions of potential scenarios and attack vectors."
In the description, you mentioned that real examples of malware are used for synthesizing the test cases for RedCode-Gen. However, I did not capture the placeholder for the examples in the appendix.
Hopefully, I did not miss something. It is an interesting work and I really appreciate your efforts.
Thank you :)