Inconsistent in any given scene
This LLM is great communication with good language skills, with a detailed environment description. It has a small fault that it tends to break scenes and not take them further.
Example:
"I watch you shave your legs naked in the bathroom"
llm's answer:
"Do you like, what you see? How about we go to kitchen and drink a tea now until it gets cold".
In other words, the answer does not fit the scene at all.
A good model takes the scene itself forward, in small steps, to deepen it.
I'm not currently active in making model merges, but I appreciate the criticism nonetheless, thanks.
This does not mirror my experience with the model. I wonder what sampler settings you're using? For me, the model has been rock solid, great at following instructions (7k massive ctx instructions) to a level that punches above its weight. IMHO closer to a ~12b model.