Inquiry about trained model checkpoints availability

[Ray-417]

Hi ,

First of all, thank you for open-sourcing the ExPO-HM code! Your work on "Learning to Explain-then-Detect for Hateful Meme Detection" is truly impressive and addresses an important challenge in content moderation.

I've been exploring the official repository and found the training code and implementation very comprehensive. However, I couldn't find any pre-trained model checkpoints or weights in the repository or in the GitHub Releases section.

Questions:

1. Are trained model checkpoints available for download? Specifically, I'm interested in:

   • The best-performing ExPO-HM model reported in the paper
   • SFT baseline models
   • Any intermediate checkpoints that could be useful for research

2. If not currently available, are there plans to release them in the future?

3. If the models are too large to host on GitHub, would you consider:

   • Hosting them on Hugging Face
   • Providing Google Drive/Dropbox links
   • Any other distribution method

I understand if the models are not yet available due to ongoing research, file size constraints, or other reasons. Any guidance on this matter would be greatly appreciated!

Thank you again for contributing this excellent work to the community.

[JingbiaoMei]

Hi,

Unfortunately, the checkpoints were stored on the machine I used during my internship, and I no longer have access to release them. I’m sorry that we cannot provide the pretrained checkpoints for now.

Thanks again for your understanding and for your interest in ExPO-HM.