Would love to compare this model with the abliterated one
#1
by
jlopez-dl
- opened
This model:
failspy/Smaug-Llama-3-70B-Instruct-abliterated-v3
Uses this method:
https://www.alignmentforum.org/posts/jGuXSZgv6qfdhMCuJ/refusal-in-llms-is-mediated-by-a-single-direction
And I'd love to test the difference between this 6-0bpw and the abliterated one. Gently asking for the quantization of this abliterated model if it was not in your list yet.