This is a really uncensored version of
llamabackup/Llama-3.3-8B-Instruct-128K "128k context" created with Heretic
https://github.com/p-e-w/heretic
initial Refusals 93/100
-> now 7 Refusals with KL=0.08
I added 30 more refusal_markers, so there may be fewer without them.
I also added 17 more bad prompts to the dataset.
Note: This heretic model is highly uncensored; thus use it with extreme caution and care.
better than all other uncesored versions from others for this model (21.FEB 26)
