LongLM really has great potential.

[lh0x00]

I'm applying this to the Ghost 8B Beta (128k) chat version online here and it seems to work.
In general, I have not yet fine-tuned and tested the parameters against the original model (even the current version is online) but I have actually noticed that the context is long but still ensures very good quality, for example here.

This is a quick issue to share this joy with your research team. Thank you very much~

[Mooler0410]

Thank you for sharing your experience with the Ghost 8B Beta model! ! !

It’s wonderful to hear it's performing well. Your feedback helps drive meaningful research impacts!

The team