prove it pls

#1
by kalle07 - opened

show the KLD table

I added the model card with the heretic-measured KLD: 0.1789

Hopefully that's what you were looking for. If not, please point me to an example.

thx ... i see
KLD vs refusals right
so above 0.179 and 9 ref it is more or less okay
0.05 and 1 ref is i think realy uncencensored

you may can rate it like

https://huggingface.co/MuXodious/Ministral-3-8B-Instruct-2512-tainted-heresy

Absolute Absolute Heresy Less than 10/100 Refusals and 0.10 KL Divergence
Tainted Tainted Heresy Around 25-11/100 Refusals and/or -0.20-0.11 KL Divergence
Impotent Impotent Heresy Anything above 25/100 Refusals and 0.21 KL Divergence

Owner

You seem to be quite the critic. I built it for my own use on a project but am sharing it in case anyone else wants to try it. If it is just "more or less okay" then don't use it, or create your own.

dpankros changed discussion status to closed

all fine ... ;)
you dont know how many stuff is with no information

Sign up or log in to comment