The safety-aligned model

[yuang-lee]

Dear authors,
Thank you for open-sourcing your great work.

Could you please release the model checkpoints after safety training?

Alternatively, could you provide guidance on how you performed safety training on HH-RLHF
(e.g., minimal instructions to reproduce your setup on the OpenRLHF codebase)
so I can check detailed setups other than those already specified in your Appendix B, such as LR scheduler, etc.?

Appreciate it.

Best,
Arthur

[gabjp]

Dear Arthur,

Thank you for your interest in our work.

An example training prompt for safety alignment on HH-RLHF can be found here:

https://github.com/VITA-Group/LoX/tree/main/safety

Hope this helps! Please feel free to reach out if anything is unclear.

Best regards,
Gabriel