One of my favorites for character and personality reinforcmeent
I wanted to first thank you for your work on this. Very solid model.
I spent a while trying to devise a SillyTavern prompt that would allow it to roleplay without bleeding the character response into its reasoning, and discovered that establishing specific instructions (e.g. "Analyze and survey the scene","Consider user's input","Speculate how char would respond") in a bulleted list format within the tags helped it tremendously. Figured I would share.
(edit: This post was meant for Violet_Magcap-12B. Sorry!)
Glad you are enjoying it! But this model wasn't trained implicitly on reasoning. Magmell Reasoner, Irixxed-Magcap, Violet_Magcap, or any of the models i have with GRPO in the name have reasoning training.
Oh whoops. I meant to comment on Violet_Magcap, not this one. My bad ^^