quick context: LLMs ship with safety alignment, they refuse certain prompts
the technique to remove this is called abliteration: find the refusal direction inside the models weights, then mathematically ERASE it
problem? doing it well by hand was slow, inconsistent, and required deep ML knowledge
@BirdYoru@Gabaniniboy@Arleyrev@Blender bro you was prolly gonna ask the same retarded question the other dude did huh, go watch new jujutsu kaisen season bro take some time off twitter
@BirdYoru@Gabaniniboy@Arleyrev@Blender "I'd be sad for a woman who'd fall for you" you should say I feel, but you are used to interacting with kids online so I am not surprised you are projecting this 16 year-old energy here
@BirdYoru@Gabaniniboy@Arleyrev@Blender all you do is throw insults at a caricature you try to draw yet you miss in every aspect, you are the same retard you try to run from
@Gabaniniboy@Arleyrev@Blender Search engines exist, blog posts exist, youtube exists at the end of the day even AI chatbots are okay for retarded questions, yet here you are