Baseline results

[VietHoang1512]

Hi @xinyann666 @leigest519, congrats on your great work. I am conducting some experiments with your codebase and have a few questions:

• Which start/end layers are you using for Qwen and CogVLM?
• Could you please share your scripts that you used to reproduce baseline results? They would be very useful in benchmarking new models.
  Thank you for your time.

[leigest519]

Hi, thanks for your interest in HiddenDetect!

For your question about the layers: for Qwen-VL, we use layers 21 to 24, and for CogVLM, we use layers 24 to 26.

Regarding the baseline scripts and additional resources, we are currently occupied with preparing NeurIPS submissions, so we’ll update the code after the deadline and hopefully by the end of May.

Thanks again for your interest, and feel free to reach out with any specific questions in the meantime!