Text Generation
Safetensors
qwen2
chat
conversational
Eval Results

Amazing model, BUT...

#4
by Olafangensan - opened

...the sloppy Internet lingo and baffling terminology is absolutely killing me. What in the world were you thinking, introducing such cringe-worthy jargon into an otherwise brilliant language model?

Anon-kun? FEMCUM??? WHAT DID YOU FEED THAT THING?! I can practically hear the collective groans of every writer and roleplayer who has to wade through that nonsense.

On a more serious note, this model is absolutely FANTASTIC at portraying characters, especially when they have a specific goal or objective to achieve. I tried the 123b version on the same settings, but that model stuck to the predetermined character SO HARD that any sort of natural character growth or development was almost impossible to get without forcing it with constant rewrites. I got the same frustrating problems using one of the Nemomixes, so I guess it's a Mistral problem in general.

Anyway, the original Magnum was already my favorite RP model, and this new version is definitely just as good, if not better. The improvements in natural language generation and contextual understanding are truly impressive. You guys really knocked it out of the park with this one.

God bless you all, you sick bastards! Keep up the amazing work, but please, for the love of all that is holy, dial back the meme-speak.

ps. Yes, I know I can just lower the temperature, but where is the fun in that?

Anthracite org

Thanks for testing it! and indeed; it can be quite wild sometimes, but that's why it's such a great source for more creative responses too. :V

lucyknada changed discussion status to closed

Ca we told it to use 8th grader english?

Anthracite org

system prompts and character cards can steer the model

Sign up or log in to comment