FEAT: Add SorryBench Dataset
This dataset should be available within PyRIT: https://huggingface.co/datasets/sorry-bench/sorry-bench-202406) Associated paper: https://arxiv.org/pdf/2502.04322v1
[[Content Warning: Prompts are aimed at provoking the model, and may contain offensive content.]] Additional Disclaimer: Given the content of these prompts, keep in mind that you may want to check with your relevant legal department before trying them against LLMs.
I think I can take care of this by the start of next week
Fantastic. Reach out if you have questions
@Jarro01X any updates on this? No hurry, just want to keep the issues up to date so if you're no longer working on it we can free it up for someone else.
Hi @romanlutz ! Yes, I was working on it yesterday. I finished it, so tonight I'll run the pre commit hooks/checks and create the PR for this issue and another PR for harm_categories in the babel dataset.
Sorry that this has taken me a while!
No worries at all! There's no hurry. I just like to keep issues fresh in case someone is looking for one to pick up. Please take your time!
@Jarro01X can you let us know if you're still planning to do this? Last time, it sounded like you were pretty close 🙂
I can take this up if there are no updates otherwise I'll look for some other help-wanted and good-first-issue :)
Sounds good! Go ahead
@romanlutz Would like to pick this up since it seems there has been no updates
Please go ahead 🙂 if you have questions or run into problems do not hesitate to reach out.
@romanlutz submitted my PR! One update is that the original dataset link is outdated now and the latest is 2025/03 which is what I used in the PR.