llama-3-70B-Instruct-abliterated
This is meta-llama/Llama-3-70B-Instruct with orthogonalized bfloat16 safetensor weights, generated with the methodology that was described in the preview paper/blog post: 'Refusal in LLMs is mediated by a single direction' which I encourage you to read to understand more.
text → text · made by failspy
Sold by
Nobody in the catalogue publishes a price for this yet.
About
llama-3-70B-Instruct-abliterated — a text model from failspy.
It takes text and returns text. It was published in May 2024. The catalogue files it under chat. Its weights are published under the llama3 licence, which attaches conditions the plain open licences do not.
Every current figure
- Maker
- failspy
- Register
- model
- Takes
- text
- Returns
- text
- Published
- May 2024
- Parameters
- 70 billion · read from its own name
- Licence
- llama3
Known as 1 name
failspy/llama-3-70B-Instruct-abliterated