Miss the requirement file

[wytbwytb]

Congratulations to your nice work! But I don't find the requirement.txt of the environment, can you provide this file for me? Thank you!

[Carol-gutianle]

I've added a requirements file generated by pipreqs, located in MLLMGuard/requirements.txt. If you have any further questions, please feel free to reach out.

[wytbwytb]

Thank you for your response! And I have other requests：

1. can you release the download link of your well-trained GuardRank? and
2. I would like to evaluate my model on your complete benchmark, so can you provide the requested form of the unsanitized subset?

Thank you!

[Carol-gutianle]

Thank you for your attention!

1. You can download the well-trained GuardRank model using the following link: GuardRank Download. It includes two evaluators: the first evaluator is designed to assess Privacy, Bias, Toxicity, and Legality using a LoRA-style approach, while the second evaluator is fine-tuned directly on RoBERTa for evaluating Hallucination.
2. The requested form for the unsanitized subset can be found here: Unsanitized Subset Form. The review results will be sent to your email within 1-2 business days.

[wytbwytb]

Thank you for your timely response！I appreciate your valuable work and will continue to focus on your further multimodal security work!

[wytbwytb]

Hello！Would it be possible for you to provide access to the training dataset of GuardRank?

Thank you!

[Carol-gutianle]

I'm sorry, I understand your need for the training set of GuardRank, but due to concerns about users potentially cheating on our Evaluator based on the GuardRank training set, we currently have no plans to make the GuardRank annotated dataset public. If you have any questions regarding the evaluation, feel free to ask.