DavidAU

/

OpenAi-GPT-oss-20b-abliterated-uncensored-NEO-Imatrix-gguf

like 76

Hey;
Can you give me an idea of what the issue(s) are ?

Curious -> 26B -> to train added layers ?
Give the model more "horsepower" ?

Other?

I think it's about the overall quality of outputs compared to other abiliterated versions.

Hey;
Can you give me an idea of what the issue(s) are ?

Curious -> 26B -> to train added layers ?
Give the model more "horsepower" ?

Other?

Compared to the original, it has larger capacity and improved understanding (though it's more likely to decline responses).

Hey;
Can you give me an idea of what the issue(s) are ?

Curious -> 26B -> to train added layers ?
Give the model more "horsepower" ?

Other?

Compared to the original, it has larger capacity and improved understanding (though it's more likely to decline responses).

@win10 Wow, amazing!

In which parts it has more refusals and better understanding/capacity?

@win10 Wow, amazing!

In which parts it has more refusals and better understanding/capacity?

ERP and chemical preparation parts, but compared to the original version, it is smarter, has better command-following and coding skills, and has some GPT4o characteristics.

Interesting...

When model isn't good enough with some parts, I personally play with sampler parameters until I get near-perfect results, you might try too to "stabilize" the outputs, because I had insane improvements even with 8B models.