"This model was produced using Hirundo's behavioral-unlearning approach to reduce censorship and geopolitical bias-related behavior while preserving non-target behavior."
I guess I never actually just chat with my Qwen models so I've never noticed any censorship or biased behavior. We just talk about the work, but this sounds like a good idea for other types of uses.
"This model was produced using Hirundo's behavioral-unlearning approach to reduce censorship and geopolitical bias-related behavior while preserving non-target behavior."
I guess I never actually just chat with my Qwen models so I've never noticed any censorship or biased behavior. We just talk about the work, but this sounds like a good idea for other types of uses.
I mean Orinth-1.5 35B has been out for a while and is considered a better model in that space. Not sure why they didn't start from there.