Request

#1
by isr431 - opened

I know this might be a long shot, but I was wondering if you'd consider fine-tuning Qwen 2.5 14B? It could replace Nemo as it's quite a bit smarter and I really like its writing style (thought it does contain GPT slop and refusals). It might also be worth using this model: arcee-ai/SuperNova-Medius. I would love to hear what your thoughts!

Arli AI org

Yea a 14B fine-tune might be interesting to try out! Will see what I can do, but not making any promises for now.

Yea a 14B fine-tune might be interesting to try out! Will see what I can do, but not making any promises for now.

Are you not planning to do any LLM fine tuning on a large dataset for ERP? As a general rule, all current models are very good at portraying female characters, but an order of magnitude worse at portraying male characters. They can basically be reduced to three archetypes: the perpetually needy whiner, the log, and the burned out asshole. It feels like the dataset is 99% textual examples of female characters.
Roughly the same problem with the set of words and actions for male characters in ERP. It's easy to get a nice obedient fem-bot, but a male character with a personality..... is impossible.

Sign up or log in to comment